Skip to content

Silinx 15.26.0

Choose a tag to compare

@github-actions github-actions released this 16 Nov 08:34
· 593 commits to main since this release

What's new in 15.26.0

Digital circuits on a breadboard come into Silinx: real 74-series chips wired as on the bench, tried out, shown inside
as their datasheets draw them, graded, and put on an FPGA. Live sharing follows everything the owner does, the Boolean
algebra takes up to 16 inputs, and a design and simulation project no longer shows FPGA tools nor HDL.

In short:

  • Digital circuits on a breadboard: New Source ▸ Digital Electronic Circuit, with a wizard of eight assemblies: the
    74-series gates, MSI and sequential ICs, switches, a clock, a logic probe; tried out in its Try it, its ICs shown
    inside as their datasheets' connection diagrams, in IEEE or IEC symbols; a circuit as an FPGA project's top, graded as
    an exercise, and the new example "breadboard".
  • Live sharing: a circuit edited live part by part; followers see the owner's selection, runs, simulation, Board
    Emulator and Boolean algebra; owner and followers reconnect on their own, and a follower's work is never lost
    silently; Share Project… reworked, Short Links… lists every link.
  • GitHub: a circuit or a drawing changed on both sides is merged object by object when pulled.
  • Boolean algebra: up to 16 inputs, the Karnaugh map and Quine-McCluskey; Settings ▸ Boolean Notation (Electronics /
    Engineering too) and IEC gate symbols; a movable divider.
  • Drawing editors: wires Direct, Orthogonal or Curved and a choice of wire crossings; many I/O markers at once; an
    instance's Width (arrays, shared pins); a state's look and its outputs one per line; the HDL language in each ⋮ menu.
  • Design and simulation projects: no FPGA tools nor HDL in their menus, editors, wizards and messages; Design
    Properties chooses the Project Kind and the Top-Level Source Type.
  • Boards and devices: the Arty S7, the Zybo Z7 and the EDGE Spartan 6; AMD Vivado on the Spartan-7, the Kintex-7 and
    the Zynq-7000; the 7-series' timing no longer optimistic, Place & Route's best of 3 seeds and a smaller .bit; ISE's
    bitstreams for the Spartan-3A and Spartan-3 boards.
  • Teaching and projects: the Exercise document's Class Results…; Import Project… from a folder too; the same
    project in two tabs kept in step; Open in VS Code.

In detail:

  • The state diagram's HDL Language locked while Try it is open (#4228, #4226, #4227): the FSM editor's ⋮ menu's HDL
    Language choices (#4063) stayed enabled while Try it locks the diagram (#3793), as its other choices are not; they are
    locked with them now, a locked diagram being as a read-only one, which has none. Test: ui/fsm-try-locked, in both
    interfaces (it failed on main in the nightly CI).
  • Show inside as the datasheet's connection diagram (#4224): the gates were drawn mid-package with straight
    diagonal lines from their pins, inputs and output alike, which did not tell which pin each input came from. Each gate
    now lies beside its pins, in the half of its output's row (in the middle when its inputs come from both rows: the
    7410's first gate, the 7430), facing its output pin (left for the 7402's and the 4000s' output-first pinouts); its
    inputs nest into its input side and its output leaves from its tip, with horizontal and vertical lines that cross
    neither each other nor another gate; a pin under its gate goes round it. The same layout in IEEE shapes and IEC boxes
    (core/circuit-inside.js). The pins no gate uses are named (VCC, GND); the MSI and sequential ICs keep their pins'
    names inside. Tests: circuit-inside (new: every gate IC's geometry, both standards), symbol-standard, UI
    circuit-inside-symbols (the lines as drawn), circuit-look, in both interfaces.
  • Show inside in Settings ▸ Schematic Symbols' standard (#4220): the gates inside a gate IC were always IEC
    60617-12 boxes; they follow the setting (#4113): IEEE Std 91-1984's distinctive shapes (the default), each turned
    toward its output pin, a bubble on an inverting one, or the IEC boxes as before; an open circuit is drawn again
    when the setting changes. core/symbol-standard.js's ieeeGate draws a shape in a box. The breadboard's look is
    unchanged. Tests: symbol-standard (the shapes), UI circuit-inside-symbols (new: both standards, changed with the
    circuit open), circuit-look, in both interfaces.
  • A digital circuit one module of the project (#4216): its generated HDL holds the circuit's module and one of each
    of its ICs (<circuit>_sn74ls47…), which only the circuit instantiates; they were taken for modules of the
    project: a row under the circuit in the Design pane, an entry of the module pickers, and in the MCP server the Test
    Bench Wizard's default module to test (a test bench of the 7447 rather than of the circuit), the Constraints
    Wizard's, rename_module's, convert_module's and the schematic's modules. A circuit's HDL is known by its header
    (core/circuit-modules.js) and its ICs' modules left out of those lists (their names still taken); synthesis and
    simulation compile the file as it is. Nothing was stored in the project: one that has them is right once it opens.
    Tests: circuit-modules (new: every assembly and an older circuit, VHDL and Verilog, compiled and elaborated with no
    warnings; the MCP's tools), UI circuit-modules (new: the Design pane's rows, the Test Bench's Associate Source, no
    warnings, both interfaces).
  • The circuit-look UI test with the circuit's ⋮ title Wires (#4208): since #4168 the Digital Electronic Circuit
    editor's ⋮ button is Wires (Fios), its HDL Language gone; test/ui/circuit-look asserted the old Fios e
    linguagem
    . It asserts the new title in Portuguese and in English, in both interfaces.
  • The circuit edited live (#4203): .circ.json is one of the drawings shared as a structure of its own
    (core/collab-kinds.js, core/collab-model.js: its parts and wires by id). Its parts' ids are names (the HDL's
    ports, the labels), so a new part whose id another participant took meanwhile gets the next free one of its
    sequence (U4, never U3~x). The circuit editor is bound like the others (setShared, applyModel, busy), and the
    owner's page writes the merged circuit and makes its HDL again.
  • The circuit in the drawings' merge (#4203): .circ.json is merged by its objects on Pull (in the server editions
    and the browser's), a conflict resolved Mine or Theirs and checked as Check Circuit checks it, its HDL made again;
    the repository's info/attributes gives it git's binary merge, an older Silinx's lines brought up to date.
  • A drawing's objects both sides added under one id (#4202): the editors number new objects one after the other
    (S7, s4), so two copies that each added one gave it the same id, and the merge by objects saw a conflict whose
    Mine or Theirs dropped one. The merge now gives theirs the next free id, its references (transitions, edges, arcs,
    the initial state) and a repeated automatic name (S0, inst_4) with it: both kept, no conflict
    (core/drawing-merge.js rekeyTheirs).
  • One's own Stop Following not warned (#4201): a follower that stopped following was told, as a WARNING, that the
    connection to the relay had ended (its own close ended the session with a lost connection's reason). Its close now
    ends it as its own (core/collab-client.js: the reason "you stopped following", the event marked stopped), said in
    the console as an information line; the owner's stop and a connection lost for good still warn.
  • A followed project once following ends (#4198): an editor's page stayed editable and its changes went into a
    room nobody read, lost when the page closed. Its documents are now opened again read only (the notice in their
    toolbar); the console's line (and the presence's menu) offers Keep a Copy, which saves the project as the room has
    it (its text files, its drawings as merged, core/collab-keep.js) into the page's own workspace under a free name,
    through the zip import (untrusted), and opens it (web/js/collab-leave.js). A page that changed the project and
    kept no copy warns before its tab closes and asks before the project closes (Keep a Copy, Close Without Keeping).
    The MCP's follow_project says that following ended and keeps the same copy with keep (with stop, or once it
    ended).
  • A follower at the meeting point answers no offer (#4197): a project the owner's page hosts is found through
    public WebTorrent trackers (core/collab-meet.js, Trystero's torrent strategy), which connect every page that has the
    link to every other one. A tracker gives a new follower's offer to one random peer of the room. A follower already in
    the room answered it as the owner did, and when its answer came first, the owner's answer was lost. Trystero then
    ignored that follower's new offers at the owner's page for 23 s, and after the first two seconds a follower announces
    only every two minutes (the trackers' interval). So the 15 s search ended with the codes. This was reproduced on one
    computer, the owner in Chrome and the followers in Chrome and Firefox: a second follower failed 4 times out of 5 in
    Firefox and 1 out of 4 in Chrome. A follower's connections now come from offersOnly, an RTCPeerConnection class
    that refuses an offer (Trystero's rtcPolyfill, given by meetOwner's join with answers: false). So only the
    owner answers, and a follower's offer that reaches another follower stays outstanding for the next tracker. With
    the fix, four followers in turn (Chrome, Firefox, Chrome, Firefox) each found the owner in under a second. Tests:
    collab-meet (offersOnly: an offer refused, an answer and a rollback taken; the follower's join with it, the owner's
    without it); UI collab-meet (on the browser's own RTCPeerConnection, both interfaces).
  • The dead tracker dropped (#4197): tracker.btorrent.xyz, from Trystero's own list, resolves to 127.0.0.1, so every
    page at the meeting point failed to connect to it and logged an error each time it tried again. TRACKERS
    (core/collab-meet.js) now lists the four other trackers, and so do the pages' CSP (server/csp.js) and the follow
    dialog's line naming the public servers. Tests: collab-meet (the list; Trystero's list without it), static-editions
    (the CSP's sources), UI collab-follow-gate (the dialog's line).
  • What still uses peer-to-peer links: nothing needed two followers to connect to each other. Only the owner's page
    (meetAsHost) and a follower's (meetOwner, also used when it reconnects) join the meeting point, and a follower
    already ignored every peer but the one that proves it is the owner. Editors' edits go through the owner's room (a
    star around the owner's page, core/collab-loop.js), and the codes exchanged by hand use a connection of their own
    (web/js/collab-rtc.js), not the meeting point.
  • The truth table editor's HDL link in a module of its own (#4194): its banner ("Not in sync with …") moved as it
    is from web/js/truthtable.js to web/js/tt-link.js, its mount function under the 300 lines of #405 with room for
    #4192; no behaviour changed.
  • A followed drawing changed in place, a truth table's look followed (#4193): a follower's page applied the
    owner's change of a drawing by opening it again from its file, its editor made again with its view's defaults (a
    truth table's Karnaugh map and Quine-McCluskey's steps closed); it is now put in place, its view kept
    (web/js/project-watch.js). The truth table's SOP / POS (the map's and Quine-McCluskey's), its map and steps shown
    and the output in front are sent with the owner's selection (#4192) and followed.
  • A participant's selection in its view (#4192): each participant's awareness view carries its selection
    (core/collab-select.js: a text editor's ranges, a drawing's objects as kind:id, a truth table's look), read as
    untrusted when it arrives and capped at 4 kB, its sending coalesced (at most one view every 250 ms, the last one
    always sent) so that a selection dragged or typed stays under the relay's awareness limit (#4173). A follower's
    page, while it follows the owner, marks the owner's selection in their presence colour (web/js/editor-peer-sel.js
    for the text editors, web/js/collab-select-draw.js for the drawings), cleared when they deselect or change
    document, while browsing alone and when the sharing ends; an object deleted meanwhile is not drawn. The MCP's
    follow_project gives the owner's selection with their view (ownerView.sel, and in its text).
  • A sharing window's link, shown (#4190): Show puts the whole link in a read-only box of its own, wrapped and as
    tall as the link needs (at least as tall as the buttons), in place of the masked one-line box, which cut off a long
    link (an editors' link is over 200 characters); Hide brings the masked box back, as tall as its buttons; on a phone's
    width too (web/js/link-field.js: Share Project, Share Module, Add a User, a design's link).
  • A device primitive's row in the Design pane (#4188): an instance of a UNISIM primitive (a schematic's Device
    Primitives, or an HDL instance of BUFG, DCM_SP, RAMB16_S9…) is a leaf marked "(primitive)", named in upper case with
    the chip's icon, not "(missing)" with the error's: it is the device's, not a module missing from the project
    (core/unisim.js isUnisimPrimitive); a module that is not in the project stays "(missing)", and both are translated
    into Portuguese.
  • What a shared project's followers get of the owner's runs (#4185): the console lines, the reports and the Design
    Summary, from the page and from the MCP's run_process, have the owner's computer's names replaced by one list the
    server makes (server/owner-secrets.js, GET /api/toolchain's redact): the workspace and the exchange folder (as
    given and as the system resolves them) as <workspace> and <exchange folder>, Silinx's configuration folder, the
    home folder as ~, the ISE host and an absolute build folder there; the Design Summary is redacted too.
  • The single file's Content-Security-Policy (#4185): the assets' site by its folder (https://pmnmalo.github.io/Silinx/,
    the development site under it), never the origin every GitHub Pages site of the account shares (server/csp.js
    siteSource).
  • The paths a project's lists name, checked as the file routes' (#4182): a hidden file or folder that a project's
    silinx.json lists (.git/config) is never built nor copied into build/src/ (an error naming it), nor a test bench
    of the simulation models, nor a file an exercise reads beside it (one of its problems); Vivado's constraints file is
    one of the project's (inside its folder, no hidden one); Sync with the .xise leaves out a hidden file and
    silinx.json as an import does.
  • A bitstream named by its absolute path (#4182): resolved once as the system resolves it (its links, then its
    ..), and that path is the one its project's trust is checked on and the one programmed.
  • Delete Project (#4182): deletes a project's folder only, one with its silinx.json (a 404 for another folder of
    the workspace, as for none).
  • Vivado's settings script (#4182): checked again when the ssh build folder changes, as ISE's.
  • A hidden test in the student's report (#4178): its line is its verdict and points only, whatever its verdict (a
    Runtime Error's or a Time Limit's messages too, not only a Wrong Answer's), as docs/CLASSROOM.md says of the issues
    silinx classroom grade --issues opens; the teacher's own console (silinx grade, the Results and Exercise
    documents, the MCP's grade) keeps every test's messages (core/grade-report.js textReport's teacher).
  • The MCP server's exchange folder, checked as a project's files are (#4177): a path of it is never a hidden file or
    folder (by its name or through a link), nor one in the workspace or Silinx's configuration folder; a new file written
    there (grade's student's copy, a CSV) has its kind's extension (.json, .csv).
  • The Classroom template (#4177): its project, when the MCP's grade template makes it, is marked as made through
    the MCP server (untrusted until the user trusts it), as new_project's; a visible test's file that is one of the
    starter project's (its top, its silinx.json), a hidden one or build/'s makes no template, its error in the console
    (Classroom Template…) or the tool's result.
  • find_in_project's time bounded (#4177): a regular expression that backtracks without end stops after 5 s,
    nothing replaced, and the MCP server answers again.
  • The local server serves only the app's assets from the package folders (#4172): from /core, /packages and
    /vendor (whole package folders, on the origin that holds the server's token) it serves no document (HTML, SVG, XML,
    PDF) nor a folder's index, and every response but a script's carries Content-Security-Policy: default-src 'none'; sandbox (a script keeps none: ELK's worker runs under its own response's CSP).
  • A follower's presence rate-limited (#4173): in a live-shared room (core/collab-room.js: the server's relay and a
    room the owner's page hosts) a follower's awareness frames (who they are, what they look at) are at most 8 KB, 10 a
    second after a burst of 40; the rest are dropped and the follower is told once ("Follow a shared project: … some were
    not passed on").
  • More MCP results labelled "Untrusted content" (#4174): git_status, git_pull's files and warnings, grade's
    run, timing_paths, related_view, read_symbol, and check_drawing's and check_constraints's problems: they
    hold a project's or a remote's own names and text.
  • The update reports write only numbers and part names (#4169): the Yosys and nextpnr update pull requests' comment
    takes a measurement of the jobs that ran the new package as outside text (#3417): a cell holds a finite number or -,
    a part its name or ?; a measurement that is not a number no longer stops the report
    (scripts/yosys-update-report.mjs cleanQuality).
  • Pages after the repository's own release (#4170): pages.yml runs after a Release run only when it is this
    repository's .github/workflows/release.yml, started by a push (a workflow is matched by its name alone);
    test/workflows.test.js checks every workflow_run trigger that way.
  • The ISE runner's install warns of a personal host name (#4171): GitHub's "Set up job" step prints the machine's
    host name in every job's log, and a public repository's logs are public: scripts/ise-runner/install.sh warns when
    the host name looks personal (macOS's default forms) and says how to give it a generic one; the install goes on.
  • The digital circuit's ⋮ menu without HDL Language (#4168): the Digital Electronic Circuit editor's ⋮ menu keeps
    its Wires and Wire crossings; its HDL file's language is chosen in its Source Properties (and by the MCP's
    set_source_properties), as the Boolean algebra's (#4159). The schematic, state diagram, ASM chart and Petri net keep
    their HDL Language (#4063). Test: ui/drawing-language-menu (the circuit in Verilog and back through Source
    Properties), in both interfaces.
  • The history scan passes on main again (#4167): scripts/check-private-data.mjs --history, which CI runs at a
    release (#555) and without which the release is not published, found the made-up home folders and ISE login of the
    redaction test (test/collab-runs.test.js) in the commit that added it, placeholders since: they are allowed in
    that file, as the other values in the history only.
  • The pull requests' check waits less in the runners' queue (#4166): at most 6 parts of a pull request run at
    once (max-parallel: the parts stay planned at about 20 minutes, under their 60-minute limit, where 6 longer parts
    of the whole suite would take 58 each); a draft runs no test and fails related-tests, the tests running when it
    is marked ready (gh pr ready); and a hub (HUBS in scripts/related-tests.mjs: web/js's modules imported by 15
    page scripts or more, i18n.js, app.js, ui.js, drawings.js…) selects only its own UI test, not those of its
    importers nor those that name it (import('/js/i18n.js')), which the nightly runs: Settings ▸ Schematic Symbols'
    pull request (#4113) selected 210 UI tests in 4 parts, now 24 in 1.
  • The Digital Electronic Circuit Wizard's assemblies of ICs (#4164): after Breadboards and a supply, and Switches and
    LEDs, eight assemblies on a full breadboard and its supply, their chips placed, powered and wired to switches,
    buttons, a clock, LEDs and a display (core/circuit-assemblies.js; the list stays core/circuit-templates.js's):
    switches into a 7447 and a common anode display (its segments pulled up) or a 7448 and a common cathode one; a clock
    (1 to 50 Hz) into a 7490 (BCD) or a 74161, a 7447 and a display, a button rst clearing it; a 7483 or 74283 adder; a
    gate IC (7400, 7402, 7404, 7408, 7432, 7486), each gate on switches and an LED; a 7485 comparator; a 74138 decoder on
    8 LEDs or a 74151 multiplexer and its data switches; a 74164 or 74194 shift register with buttons clk and clr
    and a switch din; a 7474 D flip-flop with a clock button and switches on D, PRE̅ and CLR̅. Each one clean for the
    circuit's checks (no warnings). The MCP's new_source takes them (template, template_params: chip,
    display, hz). In Portuguese. Tests: circuit-assemblies (new: each one simulated against a reference written
    from the datasheets, in the engine and in its HDL, VHDL and Verilog, with no warnings), circuit-templates,
    mcp-new-source, UI circuit-wizard (both interfaces, Portuguese).
  • The Digital Electronic Circuit Wizard without Empty (#4161): its assemblies are Breadboards and a supply, and
    Switches and LEDs (core/circuit-templates.js); circuitTemplate refuses an unknown one, naming the valid ones. The
    MCP's new_source refuses template: 'empty' the same way ("template: one of boards, io"), writing nothing. New
    Source ▸ Digital Electronic Circuit without the wizard still makes a half breadboard and its supply. Tests:
    circuit-templates, mcp-new-source, UI circuit-wizard (both interfaces, Portuguese).
  • The Boolean algebra's ⋮ menu without HDL Language (#4159): the Boolean algebra editor keeps its Notation only,
    which lists VHDL and Verilog too; its HDL file's language is chosen in its Source Properties (and by the MCP's
    set_source_properties). The other drawings keep their HDL Language (#4063). Test: ui/drawing-language-menu, in
    both interfaces.
  • An expression shown in part, in core/ (Part of #4150): the bound the truth table's results show a large
    expression with (its first 10 000 operators and literals, #3939) moved from web/js/tt-results.js to
    core/expr-part.js as it was, for the MCP server's related_view; no behaviour changed.
  • Expressions shown in part, in the MCP server too (#4150): related_view's results write each expression through
    core/expr-part.js as the editor does: its first 10 000 operators and literals, then (shown in part: N literal(s) in all). A table of 16 inputs gave 353 million characters in the text notation and ran out of memory in the Electronics
    one; now 2 MB, about a second.
    • A one-input NAND or NOR (its input written twice) gives its input half of the budget left, so that nested ones stay
      within it as written: a NAND form of 16 inputs was 6.5 million characters in the editor too, now about 30 000.
  • The tests' timings refreshed (#4145): test/timings.json from the nightly CI run of 2026-11-15, the UI test files added
    since measured, so that test/test-shard.test.js passes again and the parts stay balanced.
  • "Import Project…", the label of an import that takes a folder too (#4132): the File menu's and the Start page's
    command, its Portuguese, the MCP server's import_project description, the docs and the UI tests; no change in what
    it does.
  • Part info (#4126): under a selected part's fields and pinout, its details from core/circuit-part-info.js
    (partInfo): what a part that is not an IC does (a breadboard's strips and rails, the supply, the switches, a pull, an
    LED, a display, the probe) and whether it is a port of the HDL; an IC's chip, family (74LS TTL, 4000 CMOS), package,
    supply pins, logic (each gate with its pins' names), typical delay (it replaces the note under the pinout),
    open-collector outputs (the 7447's too) and how an input left open reads (TTL 1, CMOS X, as the simulation does); each
    pin's role. Translated into Portuguese. The MCP server's circuit_parts gives them for one part (its info). Tests:
    circuit-part-info (every part, against its description), mcp-circuit, UI circuit-parts (a 4001's, a clock's, a
    7447's in Portuguese) in both interfaces.
  • The Parts panel's drawings fitted (#4124): a part that is not an IC was fitted to its pins and a guessed margin, so
    the supply, the pulls, the switches and the LED were drawn tiny in a corner of their boxes, and their texts were all
    left out. Each is now fitted to its own drawing's box (measured once), centred, a small part magnified 1.2 times at
    most; only its name is left out. UI test circuit-parts-panel (every such drawing centred and filling its box, the
    supply's and the pull-up's marks) in both interfaces.
  • CLAUDE.md brought up to date (#4123): the devices and toolchains Silinx has (the Spartan-6, iCE40, ECP5, Gowin and
    7-series boards, design and simulation projects, AMD Vivado), the examples' targets of each family, code-size's
    measured folders and its empty lists of exceptions, web-i18n scanning every web/js file, a circuit among the
    drawings and no HDL badges in a design and simulation project.
  • The contributor and technical docs brought up to date (#4123): ONBOARDING.md (the boards, programming with
    openFPGALoader and the MCP's program_board, the code map's circuits, sharing, grading, Vivado's flow and relay);
    SECURITY.md (the scope's relay/ and packages/silinx-hdl/, AMD Vivado's settings checks and trust refusal, the
    MCP's program_board under the trust gate, its untrusted results, a Silinx project imported from its folder, the
    relay's owner resume); docs/UI-GUIDELINES.md (the select arrow and add tools that stay on, a document's look as a
    choice of its ⋮ menu, Show / Hide buttons, Try it, the circuit's Parts panel, movable dividers, no HDL badges in a
    design and simulation project); docs/ARCHITECTURE.md (Vivado's flow, the project kind, circuits, the boards'
    modules, sharing, git); docs/TOOLCHAIN.md (Vivado, the back ends, the 7-series' table with nextpnr-xc7 0.3.0 and the
    breadboard, the references of the 7-series and Gowin); packages/silinx-fpga/README.md (the families, the ECP5's and
    Gowin's nextpnr); docs/HISTORY.md (the nightly CI, the milestones since the rename).
  • The user guides brought up to date (#4123): GETTING-STARTED.md and COMECAR.md (kept parallel, the Portuguese with
    the interface's own names, "Exemplo:", "Projeto de teste:") list the breadboard example, New Project's Project Kind
    and Top-Level Source Type, the circuits on a breadboard, the import of a project's folder in the browser, Share
    Project's Generate, the design and simulation project's export, AMD Vivado for the 7-series boards and Open in VS
    Code; the English says how to change the interface's language, as the Portuguese did. CIRCUITS.md: the displays and
    the LED bar are outputs, a circuit is a Top-Level Source Type, the toolbar's Show inside, ☰ Parts, ☰ Properties and
    ⋮ menu, Curved wires a check item, and a circuit not merged by the drawing merge yet (phase 4). CLASSROOM.md: the
    --issues sentence whole again. PROJECT.md: silinx.json's kind and topSourceType, synthTool 'vivado', and
    the families of /devices.
  • The MCP server's instructions and descriptions brought up to date (part of #4123): the instructions name the
    boards beyond the Basys2 and Nexys2 (list_devices), the design and simulation projects, the circuits and the drawings'
    tools (check_drawing, simulate_drawing, related_view, set_notation, convert_module, symbol_info, place_instance,
    circuit_parts, edit_circuit…), short_links and grade, and a live share with editors; share_project is no longer called
    read only (with editors its start result no longer says it either, and its description says that the interface's
    Live share always makes the editors' link, share_project only with editors: true); new_project and
    new_project_from_example take any board of list_devices; set_design_properties says its Project Kind and Top-Level
    Source Type; run_process says which 7-series devices Vivado alone implements and the Artix-7's Silinx toolchain on
    trial; new_source names the LPF, CST and XDC constraints and the circuit; circuit_parts its kinds pull and probe, its
    pins' weak, probe, oc and nc and the full breadboard's holes; simulate_drawing's title the circuits; related_view's
    notation its fsm view and engineering's AND as a point, its results their gates by type; check_syntax a drawing that
    makes no module and Check Design; the conversions the notation of their conditions. silinx help gives silinx ucf
    any board and silinx mcp its folders (SILINX_WORKSPACE, SILINX_MCP_FOLDER).
  • The README brought up to date (#4123): the boards of the Silinx toolchain (the ULX3S, the Tang Nano 9K, the
    Basys 3 and Nexys A7 too), the boards and families added (EDGE Spartan 6, the Vivado boards, the Spartan-7), the
    Boolean notations and the 16 input bits of Algebra from Module; what it did not say yet: design and simulation
    projects and the Top-Level Source Type, a Silinx project imported from its folder, arrayed instances and their
    shared pins, the I/O marker rectangle, the wire crossings, the circuit editor's ICs, Parts panel, moves and Try it bar,
    the gates by type, a state diagram's look, the sticky drawing tools, the Petri net editor, two tabs, a live share
    reconnecting, Class Results, the 7-series' best of 3 seeds and compressed .bit; the MCP tools' new arguments and
    the tools it did not list.
  • The MCP server's new_project takes the Top-Level Source Type, and edit_circuit sets a wire's colour, as the
    interface does (#4121):
    new_project takes top_source_type and language, as New Project's Top-Level Source Type
    and Preferred Language: the project starts with its top document and the top module top (a diagram's with its HDL
    module, generated from it), the files the wizard writes, made by the same function (core/top-source.js's topSourceFiles); a design and simulation project's top a drawing's. edit_circuit's set
    sets a wire's colour, through setWireProps of core/circuit-edit.js, which the wire's Properties' Colour calls
    now too (an unknown wire, colour or property refused with its message).
  • New Project's top document made in core/ (Part of #4121): the files a Top-Level Source Type's top starts with
    (#4105), moved from web/js/wizards.js to core/top-source.js (topSourceFiles) as they were, the wizard writing
    them in the same order, for the MCP server's new_project to make them too; no behaviour changed.
  • New Source: one list of the sources' extensions taken off a typed name (#4120): sourceBase (core/ident.js), used by
    the wizard's File Name page, its Summary page and the MCP server's new_source, which each had their own list:
    .circ.json was in none of them, and the Summary's missed .fsm.json, .pn.json, .lab.json, .c and .cpp too.
  • A view=try link on a circuit (#4119): core/design-link.js's TRY_DRAWING takes the circuit (.circ.json), so
    a circuit's module has a Try it link (its name, or its generated HDL's module); the link opens the circuit and starts
    its live simulation (web/js/link-open.js, through the circuit editor's toggleSim), and Share Module… on a
    circuit's row chooses Try it. The message of a module with no Try it names the circuit too (in Portuguese as well);
    the MCP server's project_link says it. Tests: design-link (the breadboard example's circuit), mcp-link
    (project_link and open_link on it), UI design-link (the link opens the circuit's Try it; Share Module…) in both
    interfaces.
  • Grading: a circuit is one of the drawings an exercise allows (#4118): circ joins the drawing kinds of
    allow.drawings (core/grade-format.js DRAWINGS), which an exercise without its own list allows, as it does the
    schematic, the state machines, the truth table and the Petri net; docs/GRADING.md's example lists it.
  • Algebra from Module…'s dialog: the real limit (#4117): its hint names the limit the module is read with, the
    editor's (MAX_EDIT_INPUTS, 16), from that constant; its Portuguese a pattern of the number.
  • Turned parts drawn right (#4114): a part was drawn from its turned pins' box with the unturned part's offsets (an
    IC's body and notch, a switch's or a button's body, the supply's, the clock's, a DIP switch's levers, a display's
    segments, a probe's pen wrong at 90° and 270°, on the wrong side at 180°). Every part is now drawn unturned and turned
    whole about its place, as its pins are, and as a breadboard already was (#4065); its texts stay upright (a breadboard's
    letters and numbers too), a part's name starting away from its body when it ends up beside it. UI test
    circuit-rotate (every kind of part at the four turns: its drawing the unturned one turned, its pins on its pins, its
    texts upright, its name clear of its body; an IC turned by the toolbar's Turn) in both interfaces.
  • Settings ▸ Schematic Symbols (#4113): a preference beside Boolean Notation (View ▸ in the ISE style interface),
    kept in the browser: IEEE Std 91-1984 (distinctive shapes), as before, or IEC 60617-12 (rectangular), a box with
    its qualifier inside it and the type under it, a negation the negation circle (a bubble), not the polarity
    indicator. The schematic editor and its read-only views, the palette's and Symbol Info's drawings, and the RTL and
    technology schematics follow it, drawn again when it changes, and so do their printing and their PNG / SVG export;
    every pin stays where it is, so the wires stay connected, and the saved schematic, its netlist, its HDL and its ISE
    export do not change (core/symbol-standard.js, its boxes' geometry tested against the distinctive shapes' pins). The
    circuit editor's Show inside keeps its IEC boxes, now with the same table of qualifiers. The MCP server's
    set_schematic_symbols (ieee, iec) sets it for its user, as set_notation the notation. Portuguese
    ("Símbolos Esquemáticos").
  • A test's follower browses alone (#4111): test/ui/readonly-notice.test.js's follower of a shared project clicks
    the Design pane first, as a user's own action, so the owner's activity (#4002: the Design Summary in front) no longer
    brings the summary back over the documents the test opens; it failed now and then on CI and locally (the 30 s wait
    added before was not the cause, removed).
  • An editor's changes written to the owner's disk in their order (#4110): each change an editor made to a text file
    was written by a write of its own, with nothing ordering them; a replacement (a deletion, an insertion) or fast typing
    could leave the file on an earlier text when an earlier write finished last, until the next change (a rare failure
    of test/mcp-collab.test.js's promote / demote). core/collab-doc.js watchDoc takes inOrder: each call once the one
    before has ended, each file's text as the shared document has it then; the interface's owner page
    (web/js/collab-share.js) and the MCP's share_project (server/mcp-collab.js) use it.
  • "Least gates" (#4106): the truth table's result and form "Least number of gates" is named "Least gates", beside
    "Least delay" (Portuguese: Menos portas); MCP related_view too.
  • The Top-Level Source Type (#4105): its types in core/top-types.js, Petri Net, Circuit, C Function (HLS) and
    C++ Function (HLS) added; New Project makes the top of each (createTopSource: the drawing, or the function top);
    Design Properties' a choice, saved as ISE's property is, nothing made, the console saying it, undone by Undo Design
    Properties; its Top Module row gone (Set as Top Module chooses it). A design and simulation project's types are the
    drawings' (no HDL, no C / C++), in New Project and Design Properties. The MCP server's set_design_properties takes
    top_source_type, alone or with the rest, with the same choices and checks.
  • Design Properties' Project Kind (#4103): the dialog has New Project's Project Kind field (web/js/wizards.js, the
    choices of web/js/project-kind-ui.js KIND_OPTIONS), the project's kind shown; "no FPGA" hides the board, the
    device, the synthesis tool and the language at once, FPGA design shows them; the conversions are #3780's, both ways,
    with their messages, and Undo Design Properties gives the kind back. The board list's "Design and simulation only
    (no FPGA)" entry is gone. The MCP server's new_project and set_design_properties take kind (fpga, design):
    kind: fpga alone makes a design project an FPGA project with no board (its default device), as the dialog's FPGA
    design with None Specified; design_only stays as its deprecated alias (false now that same conversion).
  • Wire crossings (#4101, web/js/wire-crossings.js): in the ⋮ menus of the schematic (its editor, a read-only
    schematic, the RTL and technology schematics) and of the circuit editor, the Hop-overs check box is a titled choice as
    Wires is: Wire crossings, Hop-overs (a small arc where two wires cross without connecting, the default) or
    Plain crossings (the wires straight across each other; a connection keeps its dot). The same setting, remembered
    as before; Portuguese.
  • The truth table's divider (#4099, web/js/tt-divider.js): between the table's column and the expressions, dragged
    with the mouse or a finger (as the main window's panels), ← → with the keyboard; the column's width remembered for the
    next table opened, the expressions keeping 200px; not printed.
  • The live ASM chart's and Petri net's UI tests leave the side panel's field without a click on the toolbar (#4097):
    test/ui/collab-asm.test.js and collab-pn.test.js clicked the middle of the editor's toolbar, which since its new
    tools is a few pixels from Delete selection, and with the Linux runner's fonts on it (the selected state deleted);
    the field is blurred instead.
  • AND written as a point (#4094): the Electronics / Engineering notation writes AND as a point: the truth table's
    results and groups, a state diagram's tables, the conditions a conversion writes, the notation's example in the
    menus. The middle dot is still read when typed.
  • The Boolean algebra panel's captions (#4093): no "Expression" caption over the function's field; "Results" is
    "Expressions".
  • A ticked result stays ticked (#4092): switching the Karnaugh map's or Quine-McCluskey's result between SOP and POS
    while it is ticked (the output's HDL logic) keeps it ticked, the output's logic the form clicked; unticked, the switch
    changes only what is shown.
  • The gates of a result by type (#4091, core/logic-sch.js circuitGates): every result of the truth table says its
    terms, its literals and its gates by the library's names, as its schematic draws them: a complemented input of an AND,
    OR, NAND or NOR one of its inverted inputs (AND2B1, OR3B2…, #3786), an inverter (INV) only where one is drawn (an XOR's
    complemented input, a complemented output), a gate wider than the library's the gates it is split into (AND5, AND2…).
    • The circuit's gates and delay count the same gates: an inverted input adds no gate and no level. The forms of the
      least gates and of the least delay are chosen on them: an output in one of them may be in another form than before,
      its HDL module regenerated with the same function.
    • Least number of gates and Least delay show the function itself beside their name, not the name of the form it is
      in, and under it its terms, literals and gates.
    • MCP: related_view's results say the same (each form's terms, literals and gates; the data's cost and gates).
  • The ASM editor's UI test after the encoding moved to the ⋮ menu (#4086): test/ui/diagrams.test.js expects the
    menu's Encoding choices (✓ Binary, Gray, One-hot, Enum / auto) and no select in the toolbar (#4055).
  • A wire from A to B in the circuit's style (#4083): the circuit editor's Wire tool makes its wire from the click on A
    (a pin or a hole) to the click on B in the circuit's wire style (#4080), the clicks between them the points it goes
    through: Direct straight, Curved a curve (a sag between two points), Orthogonal routed by the schematic's router
    (core/schpath.js on the circuit's grid, core/circuit-route.js routeOrthogonal, addStyledWire): never over
    another part's pin nor into an IC, along no other wire, around the breadboard where it can, the clicks its
    waypoints; the draft shown as it will be made; a boxed-in pin's leg drawn straight, said in the console. The MCP's
    edit_circuit wire does the same, its route said. Tests: circuit-route (the routes checked point by point, the
    nets kept), circuit-wires, mcp-circuit, UI circuit-wire-draw (new) in both interfaces and Portuguese.
  • The schematic's wire styles (#4081): the ⋮ menu's Curved wires tick box is now the Wires choice (Direct,
    Orthogonal, Curved; a ✓ on the current one, an undoable edit, in Try it and the read-only views too, as before).
    The schematic's wireStyle is direct, curved or none (orthogonal: older files read as before). Direct draws a
    wire straight from end to end, or through its own points (the knots a curved wire has: kept from one style to the
    other); a wire drawn by clicks keeps the points clicked; the handles bend it. Each wire keeps its rectilinear path
    (points), hidden in Direct and Curved: its connections, its net, Check Schematic, the VHDL / Verilog, Rearrange
    Wires / Schematic and ISE's .sch (written orthogonal, as before for curves) are the same whatever the style.
    The ids and labels are in core/wire-style.js, shared with the circuit editor (#4080). The MCP's new
    set_wire_style tool (style: direct, orthogonal or curved) sets it. Portuguese: Fios, Direto, Ortogonal,
    Curvo.
  • The circuit's wire styles (#4080): the ⋮ menu's Curved wires tick box becomes a choice, Wires: Direct (a
    straight line from point to point, the default, today's wires), Orthogonal (horizontal and vertical segments) and
    Curved; the circuit's wireStyle (direct, orthogonal, curved; direct not written), an edit undone by Undo; the
    MCP's edit_circuit { op: "wires", style } takes the three: the schematic's styles and their labels (core/wire-style.js, #4085, its default here direct),
    drawn by core/circuit-wires.js (orthoPoints).
    Tests: circuit-wires (every orthogonal step horizontal or vertical, through the same points), mcp-circuit, UI
    circuit-look in both interfaces and Portuguese.
  • The sharing windows' links (#4073): web/js/link-field.js makes a link only when its Generate is clicked
    (the full link, the short link: the same gist for the same expiry; the live viewers' and editors' links, each by its
    own, kept for the share's later windows; Add a User…'s); a choice changed clears it. The buttons in the order
    Generate, Copy, Show, Show QR Code; the masked box stretched to their height. The live window shows the sharer's name
    once shared (not a field) and has no Add a User… button (the Sharing menu's, whose Copy Link is Copy View Link). A
    viewer's link is …#view=<room>.<key> (core/collab-link.js; #follow=, the links made before, still opens), for
    the MCP's share_project and follow_project too.
  • A circuit's schematic (#4069): the badge [Schematic] of a circuit's Design row opens a read-only schematic
    generated from the circuit (core/circuit-sch.js: its HDL in its logic form, the digital model synthesis reads, drawn
    by the HDL-to-schematic conversion: the gate ICs' gates as gates, the other ICs as blocks, its switches and LEDs as
    I/O markers; web/js/drawings-circ-sch.js), its banner's link to the circuit, redrawn when the circuit changes (out
    of sync, its errors said, while the circuit has errors); the MCP's related_view schematic of a .circ.json.
    Tests: circuit-sch (the half adder's schematic's own HDL simulated against the half adder, the breadboard example's
    blocks), mcp-views, UI circuit-schematic (new) and circuit, in both interfaces and Portuguese.
  • The circuit's Try it bar (#4067): while trying a circuit, a green bar above the canvas (as the schematic's): Try it,
    its Reset (the simulation again from time 0) and its Stop (Esc too); the time in the status line. The toolbar's Reset
    goes. UI tests circuit (Reset, Esc, Portuguese) and try-it-body (the bar's green, Stop) in both interfaces.
  • No circuit waveforms of their own (#4068): the Waveforms button, the Try it replayed in SSim
    (core/circuit-replay.js, web/js/circuit-waves.js) and simulate_drawing's waveforms option go; a circuit's
    waveforms are its test bench's, simulated as any module's. Tests: mcp-circuit (the option refused),
    circuit-probe.
  • A drawing's HDL language in its ⋮ menu (#4063): an HDL Language choice in the ⋮ menu of the schematic,
    state diagram, ASM chart and Petri net editors (web/js/drawing-lang-menu.js): Preferred Language
    (the drawing follows the project's), VHDL or Verilog, the ✓ on the drawing's; the same choice as Source Properties'
    (setSourceLanguage, the project's sourceLanguage, the MCP's set_source_properties): the HDL file made again in
    that language. None in a design project nor in a read-only view. The circuit's Properties lose their read-only HDL
    language field; an open circuit follows a language change. Tests: UI drawing-language-menu (new; every drawing,
    the circuit in Verilog and back, a design project, Portuguese) and the editors' menus' tests, in both interfaces.
  • The circuit editor's Parts panel (#4062): on the left of the canvas (web/js/circuit-parts-panel.js): a search,
    the parts by category (folded and opened by their titles, their counts), each part a row with its drawing (an IC's
    the schematic editor's symbol of its gate, nand2, inv…, or of its block, decoder, counter…), foldable to a
    strip (remembered) and, on a narrow screen, over the canvas behind the toolbar's ☰ Parts. The Properties on the right
    keep the selection's and the circuit's. UI tests circuit-parts-panel (new), circuit, circuit-parts, in both
    interfaces and Portuguese.
  • The Digital Electronic Circuit Wizard (#4061): New Source ▸ Digital Electronic Circuit (renamed, after Boolean
    Algebra) ▸ Wizard >: an assembly (core/circuit-templates.js, the template page's): N breadboards (half or full) and a
    supply, every board's rails joined to the next one's; a half breadboard with N switches sw[k] and M LEDs led[k];
    the circuit made and opened (web/js/circuit-wizard.js). The MCP's new_source takes the same (template,
    template_params). The Top-Level Source Type's Circuit (#4105) is named Digital Electronic Circuit too. Tests: circuit-templates (every board powered,
    checked by the circuit's nets and checks), mcp-new-source, UI circuit-wizard (new), new-source,
    new-source-paths, design-only, projects, circuit, design-properties-top-type, in both interfaces and Portuguese.
  • Show inside (#4060): a tick box of the circuit editor's toolbar (remembered): each breadboard's strips under its
    holes, each gate IC's gates (IEC boxes joined to its pins, a bubble on an inverting one), another IC's pins named
    inside it (web/js/circuit-draw-inside.js).
  • Curved wires and Hop-overs (#4064): the circuit editor's ⋮ menu: Curved wires (the circuit's wireStyle, an edit,
    undone by Undo; the MCP's edit_circuit wires) and Hop-overs (remembered, on by default); core/circuit-wires.js
    (the curves, the crossings, the arcs). The toolbar's options moved to web/js/circuit-options.js. Tests:
    circuit-wires (the crossings against ones reckoned by hand), mcp-circuit, UI circuit-look in both interfaces
    and Portuguese.
  • Moves that keep the connections (#4059): a breadboard moved carries the parts with a pin in its holes and the wires
    between them; the toolbar's Keep connections (as the schematic's, remembered, on by default) has the ends of the
    other wires on the moved pins and holes follow them, shown while dragging. core/circuit-edit.js movePart and
    rotatePart ({ keep }), the MCP server's edit_circuit move and rotate (keep: false). Tests:
    test/circuit-edit.test.js (the connections against the circuit's nets before the edit), test/mcp-circuit.test.js,
    UI circuit-moves.
  • Every part turned (#4065): a breadboard too, its holes and the parts on it with it, drawn turned.
  • Copy and paste (#4066): Ctrl+C / Ctrl+X / Ctrl+V and Edit ▸ Cut, Copy, Paste copy the selected part (a breadboard
    with its parts and the wires between them), pasted 2 steps right and down, further each time, new ids given, a label
    another part has left out (copyParts, pasteParts; edit_circuit's copy).
  • The Electronics / Engineering notation's bars drawn (#4058, web/js/overbar.js): the truth table's results,
    Quine-McCluskey's steps, the Karnaugh map's groups and a state diagram's transition table draw each NOT as a line over
    its group, the groups nested, each bar above the ones under it (the combining overlines, U+0305, which fonts stacked
    badly). A long barred group is not broken across lines: its line scrolls. The text is unchanged: copied, it is the
    same expression with its overlines, which the parsers read back. No new dependency.
  • The Karnaugh map under its result (#4057): the map (its order, 3D, minimal solutions and groups' legend) is no
    longer above the results: it is drawn under the Karnaugh map result while its Show K-map button is pressed (Hide K-map
    hides it), hidden when a table is opened, each table its own. Its groups are the result's SOP or POS, as clicked: the
    SOP groups / POS groups radio buttons are gone.
  • Quine-McCluskey's Show Steps (#4057): its steps are shown by a button, Show Steps / Hide Steps, not a check box.
  • The ASM chart's state encoding in the ⋮ menu (#4055): the toolbar's "Encoding:" selector is gone; the ⋮ menu has
    its Encoding title, then Binary, Gray, One-hot and Enum / auto, a ✓ on the chart's (menuChoice, as the state
    diagram's, #3289), moved by Undo and Redo; the encoding stays in the status line (#2763). Test: UI
    asm-editor-chrome in both interfaces and Portuguese.
  • An arrayed instance's Shared pins, its inputs alone (#4053): the Shared pins of its Properties list only its
    input pins, an output being driven by every copy; the MCP server's shared refuses an output or a bidirectional pin,
    naming it; a mark left on a pin that is not an input (a file written before) is ignored, the pin taking the rule by
    its net's width (core/sch-array.js isSharedPin, sharedList).
  • The circuit editor's Try it body (#4049): while a circuit runs (Try it), its Properties panel is the Try it's body,
    the lighter green of #3924 (try-it-body), plain again once it stops; test/ui/try-it-body.test.js checks it with the
    other editors, in both themes, its text AA.
  • No false "mode IN can not be updated" / "mode OUT can not be read" (#4048): with an instance's output and one of its
    inputs on one net, the elaborator makes them one signal, and the design checks took a write of the output for an
    update of the input, and a read of the input for a read of the output. The rules now look at the port the statement
    names (core/lint.js); a port written or read by its own name is still an error (VHDL's case-insensitive names too).
    The breadboard example's simulation model has no design-check message now.
  • The interface's Portuguese labels, one line per module (#4041): web/js/i18n.js merges each label module (the
    editors', the wizards', the labs'…) by a helper, keep(strings, patterns), one line each instead of two, in the same
    order: the same labels and patterns (the file at 983 lines, under the limit of #405, for the modules to come).
  • The breadboard example in the comparison with Vivado (#4035): recorded on the Basys 3 and the Nexys A7
    (test/xc7-vivado-comparison.json, docs/TOOLCHAIN.md's table). Vivado builds it with no warning and its DRC clean, and
    takes Silinx's netlist with no error (3 LUTs, 267 MHz); Silinx's open flow 4 LUTs, 241 and 250 MHz, Vivado's own 4 LUTs,
    450 and 446 MHz. scripts/xc7-vivado-compare.mjs --update writes the results with the file's 2-space indentation.
  • test/ui/xc7-boards.test.js (#4030): it expects the four experimental 7-series boards of Design Properties (the
    Basys 3, the Nexys A7, the Arty S7 and the Zybo Z7, #3895), as test/ui/xc7-vivado-families.test.js does.
  • Open in VS Code (#4017): the server runs the editor (code <project folder> --goto <file>:<line>, found on
    PATH or where it is installed, spawned without a shell), or the page opens its vscode://file/… link when it is
    not found; a file of the project only (its path confined, no hidden file or folder, not a design project's HDL),
    never a program of the request's, refused to a page on another computer (GET /api/editor, POST /api/editor/open). config.json's editor (file only): code by default, code-insiders, codium, cursor, a
    program's absolute path, or off (SILINX_EDITOR=off too). The MCP server's open_in_editor does the same. None
    in the browser edition. The UI tests' server has it off (no test starts the user's editor).
  • The schematic editor's demo page loads again (#4011): the Properties panel (web/js/sch-props.js) loads the console
    only when its Width field reports an error, since #3950's static import pulled in the app and CodeMirror, which
    sch-editor-demo.html does not have (test/ui/demo-pages.test.js).
  • An old interface name out of a test's comment (#4008): test/ui/import-folder.test.js (#3965) said "Xilinx
    ISE one", which the old-name test (test/interface-ids.test.js, #2451) refused on main; it says "an ISE project's".
  • No line about an earlier version's project bundle (#4003): Import Project (.zip)…'s Silinx choice in the
    browser edition drops "It also opens a project bundle (.silinx.json) made with an earlier version."; an old
    .silinx.json bundle still opens through the window (#3548).
  • The owner's activity (part 1 of #4002): a new sealed message of the room's protocol, KIND.activity
    (core/collab-activity.js), carries what the owner does, by channel. The owner keeps each channel's latest value, sends
    it when it changes, and sends it again to a follower that joins or rejoins. A value is at most 1 MB and there are at
    most 16 channels. The room forwards no follower's or editor's activity (read only), and a page reads only the
    owner's.
  • Its first two channels (#4002, core/collab-channels.js): ui is where the owner is (the open documents, the one
    in front, the split, the left pane's page, the Design view and the task, as a design link's view, checked as one);
    design is the Design pane's selection and its rows expanded or collapsed.
  • A viewer follows them (#4002, web/js/collab-activity.js): its page opens the owner's documents, and closes the ones
    the owner closes among those it opened. It brings the same one to the front, shows the same view, page and task, and
    selects and expands the same rows. The viewer's own click on the Design pane, its pages, its views or the tasks stops
    following, as bringing another document to the front already did; the console now says "Browsing alone: is
    no longer followed".
  • MCP (#4002): follow_project gives the owner's activity (ownerActivity, and its lines in the text).
  • The owner's Board Emulators (part 4 of #4002, core/collab-emu.js, web/js/collab-emu.js): the last channel of the
    owner's activity, emu, gives each open emulator (at most 4) its board's switches and buttons, on or not, in the
    board's order, its power, whether it runs, and its clock's rate. It is sent at once when the owner presses or lets go
    of one, so a short press is not missed. The viewer's page gives its own emulator of the same design the same power,
    inputs, rate and run through the board's own controls (a switch clicked, a button kept pressed or let go as
    Shift+click does, the clock's list); its time is its own, and the design's outputs follow the inputs. A click on the
    viewer's board or its toolbar is browsing alone. The emulator's API gives its inputs, whether it runs and its rate.
  • The sharing windows (#4002): Share Project…'s live window, Add a User…'s viewer note and the follow window say
    what a viewer follows: the files and drawings, and the owner's documents, views, selection, simulations, emulators
    and runs; of the builds, only the reports the owner opens. In Portuguese too. README and SECURITY.md describe the
    owner's activity.
  • MCP (#4002): follow_project gives the owner's Board Emulators (ownerActivity.emu, and a line for each in its
    text: powered or not, running and its rate, the switches and buttons on).
  • The owner's runs (part 2 of #4002, core/collab-runs.js, web/js/collab-runs.js): two more channels of the owner's
    activity. runs carries each process's mark, the console's lines of the run going on or of the last one (its last
    1000, at most 256 KB, sent at most once a second while it runs) and the Design Summary's data. reports carries the
    texts of the build's reports the owner has open (a report over 512 KB is left out, said when the viewer opens it).
    The owner's home folder and the ISE host it reaches over SSH are replaced in what is sent, as the MCP's
    run_process already does.
  • A viewer follows them (#4002): its Processes show the owner's marks, its console the run's lines (each line
    once, from where it was), its Design Summary the owner's, and the report the owner has in front opens there with
    the same text. A follower's page now has the device database (Silinx's own data, not the server's), so the Design
    Summary names the board and shows the device's resources. The follower's own browsing is now told apart only once
    the followed project has opened, so the documents the project's opening opens no longer stop following.
  • MCP (#4002): a project shared with share_project sends the runs run_process makes of it (each step's mark,
    the end of the log, the Design Summary), and follow_project gives the owner's runs (ownerActivity.runs, with
    their marks and the end of the log in its text).
  • The owner's simulations (part 3 of #4002, core/collab-sim.js, web/js/collab-sim.js): a channel of the owner's
    activity, sim, gives each open SSim (at most 8) the time it has reached, a key of its drawn stimuli, its forces and
    its wave view (the signals shown, their radix and groups, the cursor, the markers, the zoom, the run time, the
    instance shown; not the panels' sizes). The viewer's page runs its own SSim of the same files to the same time (the
    simulator is deterministic, so the waves are the same), runs it again from 0 when the forces change or the owner
    restarted, runs the simulation again when the drawn stimuli change, and then puts the owner's view on it. A click, a
    key or the wheel in the viewer's SSim is browsing alone. SSim's API gives the time reached (now).
  • MCP (#4002): follow_project gives the owner's simulations (ownerActivity.sim, and a line for each in its
    text: the time, the cursor, the signals shown, the forces).
  • The Add I/O Marker tool's rectangle (#4000): each marker takes its pin's direction and width (a bus pin:
    in1(7:0)), turned the way the pin points; a wire end takes the direction of the pins on its net, or the net's name
    when it has one. The default names take the first number no marker, net name or instance uses, in any letter case
    (a gap is filled first, as for one marker). A wire end whose direction cannot be known (no symbol's pin on its net),
    or whose net already has a marker, is skipped. The console says how many markers were added, and which ends were
    skipped and why. The whole rectangle is one Undo step. The logic is in core/schmarker-many.js.
  • MCP: add_markers adds the markers to every open pin of an instance or of a region, with the same names and
    messages.
  • A participant's role changed by the owner (#3997): Make an Editor sends the room's signing key to that
    participant alone (a new sealed message, KIND.role, in a frame for one participant); Make a Viewer renews
    the room's signing key: the document's and the room's public key the new one (the owner's rekey, core/collab-room.js),
    the other editors given it (and what of their changes the owner may lack sent again, signed with it), the demoted
    one a viewer again; the old key's edits are refused by the room and every participant, so the old editors' link only
    follows; Share Project… shows the new one. The participant's page says it in the console and opens its documents
    again, editable or read only. A follower is a viewer in the menu, the console and the MCP (peers()' role
    viewer). The MCP's share_project has promote and demote (user: the participant's name; a demotion answers
    the renewed editors' link), and follow_project follows a renewed editors' link as a participant of its own.
  • Share Project…'s live form (#3996): one window (web/js/collab-share.js): the name field focused (it was the
    Share: row's radio button), an empty name said beside it; Start Sharing keeps the window open and shows the two
    links there, each a masked field (web/js/link-field.js) with Show, Copy and Show QR Code, no small QR code;
    Stop Sharing keeps it too, to share again; Close. The form's Copy Link and Copy Editor Link buttons are gone (the
    toolbar's Sharing menu keeps them). Add a User… (web/js/collab-users.js) replaces Add a Follower…: Viewer or
    Editor chosen, that role's link given, hidden the same way; a share the page hosts (#3227) takes the user's code by
    Take a Code…. The viewers' and editors' links are different: the editors' carries the room's signing key, which the
    viewers' has not (a test of the MCP's share asserts it). The MCP's share_project with qr gives the editors'
    link's QR code too (editQr).
  • Share Project…'s copy form (#3995): the window 720 pixels wide; its explanation ("A link that opens …") at its top;
    the Share: row's two ways side by side, each one's explanation on the line below its button (web/js/share-how.js);
    the Board row and the Lab selector gone (a link that opens a lab is the MCP's project_link lab's); the current
    view's check box labelled View:; a space between Expires: and the link. The link is a masked field
    (web/js/link-field.js, for the live links too): Show / Hide, Copy (said in the console, as Copy Link was), Show
    QR Code
    (the large one, disabled when the link is too long, its reason its tip); a short link is made by Generate,
    above Show QR Code, which stays disabled until then. A short link chosen, the window says the account signed in and
    the project's repository, its origin, private, public or not seen with this sign-in (core/gist-link.js gistAccount,
    /api/gist/account, the browser edition's sign-in); the MCP's project_link short says it too (data.account). The
    buttons Copy Link and Cancel give way to Close; the window resolves to the link last copied.
  • Short Links…: every link listed, a header check box (#3994): the table is in the window itself, which scrolls as a
    whole when the links are more than it holds (a box of 360 pixels before, about twelve rows, three on a small
    screen); the header's check box selects every link or none, indeterminate when some are. The MCP's
    delete_short_link takes all: true, every short link of the account deleted.
  • The Symbol Editor's "Edit Symbol:" plain text (#3991): its toolbar's label in the toolbar's own colour, not
    bold, with no dot and a colon after it, no longer the schematic's green Try it badge.
  • A full breadboard's holes in the message (#3974): a place that is not one of a full breadboard's holes is refused
    with its 63 columns (a1 to j63), no longer the half breadboard's 30.
  • A page's writes told to the project's other pages (#3972): each page names itself in its requests
    (X-Silinx-Client), and the local server tells what a page's request changed to the pages that have the project
    open, but not to that page itself. A request that names no page, and a run, are told to none, as before. The
    messages say "in another tab" rather than "outside Silinx".
  • The two tabs' saves at once (#3972): a page's request that changes a project waits, for 2 s at most, until
    another page's request on that project has ended and what it changed has been read; so each tab's changes are told
    to the other, and both show the file as it is on disk (before, two requests at once were told to neither).
  • A select arrow and add tools that stay on, in the state diagram and ASM chart editors (#3971): the arrow is the
    toolbar's first icon and the default tool; an add tool stays pressed (.active, aria-pressed, a crosshair on the
    drawing) and adds an element at each click on the drawing's background, until another tool, the arrow or Escape is
    picked (web/js/editor-tools.js, shared with the Petri net editor). In the ASM chart a click on a box with an add tool
    on adds the new box on that box's free exit (what selecting the box and clicking the tool did); a box with no free
    exit says so in the console. A transition is still drawn from a state's rim, and a port still draws an arrow, with
    any tool; Try it switches the state diagram back to the arrow. The MCP server is unchanged: the tools are the view's
    (its own tools add states and boxes).
  • Git's presence asked again after a failed request (#3968): a failed GET /api/git (no token yet, the server not
    answering yet) was kept as "no git" until the page was reloaded, and the Start page's git group, the File menu's
    Open from Repository… and Git submenu, and the Design pane's git entries stayed away; it is now no answer: asked
    again a few times at start-up, each time the Start page is drawn (with its projects) and when a project opens.
    SILINX_GIT=off still has no group.
  • An arrayed instance's pins marked shared (#3966): the Shared pins of an arrayed instance's Properties, a tick
    box a pin, stored in its params.shared (matched whatever the case; undone with Undo). A ticked pin is never sliced
    over the copies: drawn as a single pin, each copy given the whole net in the generated VHDL and Verilog; on a net of
    another width an error of its own (no width mismatch), and a ticked output an error, each copy driving the net
    (core/sch-array.js sharedPins, sliceArrayed, arrayedError). An unticked pin keeps the rule by its net's width.
    The MCP server's place_instance and set_instance_width take a shared list of pins (sharedList).
  • A Silinx project imported from its folder (#3965): Import Project (.zip)…'s Silinx choice has the Xilinx ISE
    one's rows (Project folder:, or .zip file:, Project name:); a folder is imported as the .zip of the same project
    (its files but the hidden ones and its build folder, core/project-folder.js), through the same import gate and
    messages; a folder picked preselects its software, a folder of no project says so; in Portuguese. The MCP server's
    import_project leaves a Silinx project folder's build folder out too. Export Project's Silinx hint ends "It opens
    in Silinx.".
  • A flaky UI test waits for what it reads (#3953): test/ui/new-source-no-add-box.test.js waits for the new
    drawing's HDL file in the project before it reads silinx.json, written after the drawing appears.
  • The project reopened at start-up is the tab's own (#3952): the open project is kept in the tab's sessionStorage
    as well as in localStorage, and the start-up reads the tab's first. A tab that closed its project reopens none when
    reloaded; a new tab, which has no project of its own, opens the last one used, as before.
  • An arrayed instance made with its Width (#3950): the Width field of a module instance's Properties rewrites its
    name with the range (r → r(7:0) for 8, back to r for 1), the name its one source (a rename with a range sets
    the field); a width not a whole number from 1 to 1024 an error in the console; undone with Undo
    (core/sch-array.js withArrayWidth, arrayWidth). Its symbol is drawn as the array (arrayedDef): the module's
    name with its copies, a pin sliced over the copies a bus pin of N × w bits (its thick lead, its width, its name with
    its range when it fits), a pin on a net of its own width shared by the copies and drawn as it is, an open one a bus.
    The MCP server's place_instance takes a width, and set_instance_width does what the field does.
  • The Symbol Editor's "Edit Symbol" badge and Exit button (#3943): the toolbar starts with "Edit Symbol:", plain
    text (#3991), so it is not taken for the Schematic editor; Exit, after the
    zoom group, closes its tab as the tab's × does (its changes saved first; if they cannot be, the Close prompt).
    docs/UI-GUIDELINES.md says when an editor's toolbar starts with what it is. Portuguese ("Editar Símbolo:", "Sair").
  • The interface names' rule kept (#3942): a design and simulation project's refused Xilinx ISE export says "not as
    an ISE project" (core/project-kind.js), and its test the same: its "a Xilinx ISE one" read as the ISE style
    interface's old name to test/interface-ids.test.js, which failed on main since #3917.
  • Truth tables of up to 16 inputs (#3939): the editor, a module's table, the expression wizard, the handwritten
    equations, a state diagram's tables and a table edited live (it was cut to 8 inputs) take up to 16 inputs.
    • A table of more than 256 rows scrolls in a box of its own; only its rows in view are drawn.
    • A very long expression (a NAND form's of many inputs) is shown in part, with its number of literals.
    • The Karnaugh map is drawn up to 6 inputs. Past that the console says there is none, and the minimal SOP and POS are
      Quine-McCluskey's.
  • The results' order (#3939): Least number of gates, Least delay, Karnaugh map, Quine-McCluskey, the canonical SOP
    and POS, then the XOR, NAND and NOR forms.
    • The Karnaugh map's result shows its minimal SOP or POS, whichever is clicked, and the map's groups follow.
    • The Minimal SOP and Minimal POS results are removed. The forms sop and pos stay, named Karnaugh map SOP / POS.
  • Quine-McCluskey (#3939, core/qm.js): the prime implicants by combining the implicants column by column, the
    chart, the essential primes, then the cover of the rest by branch and bound (Petrick's method's minimum).
    • It is bounded by work, never by time, so the same table always gives the same module. Past 20 000 implicants it
      uses a cover heuristic: each minterm left grown into a prime, then the redundant terms dropped.
    • A result not proven minimal is a console warning. So is a Karnaugh map search that was cut short (it was a note
      in the results).
    • Its steps, when ticked: the columns (✓ on those combined), the chart (× and the essential primes ✱), the essential
      prime implicants and the cover. A table too large for its steps shows its count of primes only.
    • Its tick box sets the new forms qsop / qpos in the HDL module and the schematic.
    • Tested: every 3-input function with don't cares and random 4-input ones against a brute-force reference; up to 6
      inputs as costly as the Karnaugh map's minimisation; random and structured functions of up to 16 inputs equal to
      their table, each term prime.
  • A table's own notation (#3939): the ⋮ menu's Notation is "As in the Settings" by default, else one of the five,
    saved with the table (notation). The results' own notation menu is removed.
  • MCP (#3939): related_view's results list the Karnaugh map's SOP and POS (up to 6 inputs) and Quine-McCluskey's,
    with their prime implicants, the essential ones, whether proven minimal and the steps' columns. set_output_logic
    takes qsop and qpos. set_notation with a table's project and path reads, sets or clears (settings) its own
    notation, which related_view and set_output_logic use.
  • The MCP server's Symbol Info (#3935): symbol_info (server/mcp-symbol-info.js) makes a symbol's datasheet through
    the dialog's functions (core/symdocs.js symbolDoc, core/symtables.js truthTable): the symbol by its palette,
    Xilinx library or computed name (M4_1, FDCE, AND2), its type or a project module, its parameters changing it, in
    English or Portuguese; a device primitive's resource on the project's device. In a design and simulation project
    (#3891) no HDL, no Xilinx equivalents, no device primitive nor HDL BLOCK.
  • A drawing that makes no module, told by its errors (#3934): Check Syntax (Check Design in a design and simulation
    project) and the MCP server's check_syntax say, for a drawing with errors and no HDL file that the design needs
    (its module the top, or an instance's the elaboration does not find), ERROR:Silinx - "x.fsm.json": its HDL cannot be made: …. Fix x.fsm.json … (a design project's: its module is not made), in place of "top module … not found";
    one the design does not use says nothing (core/drawing-diags.js withDrawingErrors, the same for both).
  • A text box's placeholder and tip translated (#3931): the page's translator (web/js/i18n.js) left out every
    textarea, its own attributes with its text; it now translates their placeholder, title and aria-label, as an
    input's, and leaves their text, the user's, as it is.
  • The Try it's body a lighter green (#3924): the body of the Try it (the state diagram's, ASM chart's and Petri
    net's panel; the schematic's side panel while it is tried out) in a lighter green than its bar, one class
    (try-it-body) on the theme's tokens --try-it-bar, --try-it-line and --try-it-body (dark: a darker, muted
    green); its text WCAG AA: the present state's name in a darker orange and an output at 1 on a solid green there. The
    FSM's Tables, the Petri net's Analysis and the schematic's Properties stay as they were.
  • The Board Emulator's status line in the Silinx interface (#3920): the emulator's statistics (t = … · … cycles · …
    kHz (…% of real time)) stay in its status line (statusLine(), .doc-status), shown in the Tasks interface as in the
    ISE style one, rather than moved below the board in place of the board's line (#1022); the board's zoom leaves the
    line its room. The cycles are grouped by commas whatever the browser's locale (it was toLocaleString()'s), by
    no-break spaces in Portuguese. docs/UI-GUIDELINES.md and CLAUDE.md say so. Tests: tasks (the line at the bottom, its
    look, its text in English and Portuguese, both interfaces), emulator-parts, sim-emu.
  • The notation of the Boolean expressions shown, part 1 (#3915): a preference of the viewer, kept in the browser
    (web/js/notation.js), Electronics / Engineering by default; the truth table's results, canonical forms and K-map groups
    and its own notation menu (now that setting: the five notations), and the state diagram's transition table (its
    drawing's conditions shown as typed) use it, and are drawn again when it changes. core/expr-notation.js writes a
    truth table's expressions and a condition in each notation (the overbar a combining overline, U+0305, over each
    character it covers, core/overline.js); the parsers of both read every notation back (an overbar over a name or a
    group is its NOT). The MCP tool set_notation reads or sets the MCP server's notation (config.json's notation);
    related_view's results and set_output_logic use it, or the notation a call names. Conversions (an HDL module or a
    chart made a drawing) are part 2.
  • The notation of the Boolean expressions shown, part 2 (#3915): the conversions take a notation: fsmFromHdl,
    asmFromHdl, asmToFsm and petriToFsm write the conditions they make in it (core/cond-notation.js's showCond, moved
    out of core/expr-notation.js so that they read it without an import cycle), and keep those of the previous diagram;
    with none, as before. The interface passes the Settings' (Convert to State Diagram / ASM chart, a diagram made
    again from its HDL, an ASM chart's and a Petri net's state diagrams), the MCP server set_notation's
    (convert_module, convert_drawing, related_view's state diagrams). The machine is the same whatever the notation:
    its HDL the same (the Verilog notation's ~ & | written as the bitwise operators), simulated in VHDL and Verilog.
  • The logic operators' notations in a condition (#3907): AND as a point, a middle dot or ∧ (a.b, a·b), OR as
    ∨ or + (but in a comparison's operand, where it adds: count + 1 >= max), NOT as a postfix apostrophe (a') or a
    prefix slash (/a) or ¬, besides &&, &, AND, ||, |, OR, !, ~, NOT; NOT binds first, then AND, then
    OR, comparisons tighter than AND. Bit and based literals ('1', 4'b1010) and VHDL's /= are unchanged; 1.5,
    a division (a / b) and two operands side by side (a b) are errors, which now give their column. The state
    diagram's, the ASM chart's and the Petri net's conditions share them (core/asm-expr.js); an action's + still
    adds. A transition label's / NOT (/a / z=1) is told from the outputs' /.
  • The otherwise transition (#3907): else, default, true, 1, always or no condition; kept as typed
    (no longer emptied), taken last wherever it is listed (the HDL's else), never reported as overlapping the
    others; two in one state are an error of Check (in Portuguese too). The MCP tools write_file, check_drawing and
    simulate_drawing take the same conditions.
  • A circuit as an FPGA project's top (#3901): its generated HDL has two models: the simulation's (the chips'
    delays, 'Z', 'X', the power) behind synthesis translate_off / translate_on, which XST and now the Silinx
    toolchain's front end skip (core/translate-off.js: those lines blanked, the line numbers kept), and the digital one
    (core/circuit-digital.js): the gates inlined, the MSI and sequential ICs as their entity's digital architecture,
    open collectors ANDed, a contact's level while closed, a floating net high where a 74LS input reads it and low where
    only an LED does, a 74595's three-state outputs by their enable, the clock's net its port itself. In VHDL the
    digital architecture comes first and the simulation's last (the one the simulator binds); in Verilog
    `ifdef SILINX_SIM_<NAME> chooses, its `define fenced. Gated clocks of the 74165 and the 74193 marked as
    the chips' own (no lint warning).
  • Buses (#3901): a single port's label name[k] is bit k of the vector port name (std_logic_vector,
    [w-1:0]); the circuit's check refuses a bus with a bit missing, with inputs and outputs, named as a scalar port, or
    on a part of several ports. edit_circuit says so.
  • The breadboard example (#3901): examples/breadboard, second in the list: a 74161 counting the presses of BTN0
    up to 15 (synchronised by a 74194 shift register, shown on LD4..LD7, its edge detected and the count stopped at 15 by a
    7486 and a 7408), SW0 clearing it; its self-checking test bench; built by the Silinx toolchain and by ISE on every
    board with no warning, every output of its chips read (XST warns on an open one).
  • The MSI ICs of the digital circuits (#3900, part of #3897): core/circuit-msi.js: each IC's datasheet's pinout
    (its active low pins' names barred), its delay, its model and its VHDL and Verilog bodies in one wrapper
    (core/circuit-hdl-ics.js: the inputs read as a 74LS reads them, an unknown one making every output X, the outputs
    after the delay, open collector ones a 0 or nothing); the engine and the check take an IC's model besides its gates (its
    inputs warned floating once one of its outputs is used). Tested against function tables written in the tests, through
    the engine and through the HDL simulated by Silinx in both languages, with no warnings.
  • The logic probe of the digital circuits (#3900, part of #3897): a logic probe part (no port of the HDL; the net it
    watches a signal of it), its level in the MCP's simulate_drawing's steps; a clock's ticks moving the time on by its
    half period (advance). (Its waveforms of a Try it replayed in SSim went before a release, #4068: a circuit's
    waveforms are its test bench's simulation's.)
  • The sequential ICs of the digital circuits (#3900, part of #3897): core/circuit-seq.js: each IC's datasheet's
    pinout, its delay and its model written three times the same way in the clocked style (the engine's processes run on
    their sensitivity's changes, as VHDL's; VHDL processes and Verilog always blocks), their registers X at power-on; the
    engine runs them (seq), core/circuit-hdl-ics.js writes them. The generated circuit's nets start unknown (X), not
    floating (no instant a 74LS input reads high). Tested: the pinouts, a scenario per IC (its datasheet's behaviour), and
    the engine against its VHDL and Verilog step for step on a random walk of its inputs, no warnings.
  • The circuit editor draws phase 2's parts (#3899, part of #3897): web/js/circuit-draw-parts.js (a slide switch,
    a DIP switch's levers each a click while trying it, a clock source, pull-ups, a 7-segment display's segments, an LED
    bar's lights; a full breadboard's split rails); Try it ticks each clock at its frequency in real time; the library's
    groups (the 74LS and the 4000 gates); the Properties: a switch's position as saved, a DIP switch's levers, a clock's
    frequency, a display's colour and pinout, first in the panel (then the library); Portuguese for every part's title.
  • Digital circuits, phase 2's parts (#3899, part of #3897): the full breadboard (its rails split in the middle); a
    slide switch (SPDT), DIP switches (4 and 8), pull-up and pull-down, a clock source (0.1 to 50 Hz, an input ticked in
    real time); 7-segment displays (common cathode and common anode), an LED bar; the 7401, 7402, 7403, 7405, 7406, 7410,
    7411, 7414, 7420, 7427 and 7430 (open collector outputs a 0 or nothing, a wired AND with a pull-up); the 4001, 4011,
    4030, 4069, 4070, 4071 and 4081 (a floating input reads X; their delays at 5 V). Each has its datasheet's pinout and
    its ports' names (<label>_<k>, a display's <label>_a … _dp); core/circuit-ics.js, core/circuit-io.js.
  • Their check, their HDL, their MCP tools (#3899): the check also says the open collectors on + or beside a totem
    pole output or with no pull-up, a contact shorting the supply, an input floating while a contact is open, a display's
    common on the wrong rail. The HDL writes a pulled net explicitly (the VHDL simulator's weak values: #3926).
    edit_circuit sets a DIP switch's levers and a clock's frequency; simulate_drawing takes every input by its port.
    Checked against truth tables written in the tests, through the engine and through the HDL in VHDL and Verilog. The
    editor's drawing of these parts follows.
  • Digital circuits on a breadboard, the model, the simulation and the MCP tools (#3898, part of #3897): a new
    drawing, <name>.circ.json (docs/CIRCUITS.md): real ICs (the 7400, 7404, 7408, 7432 and 7486 with their datasheets'
    pinouts and 74LS delays), a half breadboard, a supply, logic switches, push buttons and LEDs on a 0.1 inch grid, their
    nets geometric (a pin on a hole is in its strip, a wire joins its ends); an event-driven simulation with the gates'
    inertial delays, a floating 74LS input read high, an unpowered IC's outputs floating, a short X, an oscillation said;
    its check (short circuits, unpowered or reversed ICs, floating inputs, LEDs that never light, parts and wires that
    join nothing, labels); its HDL in VHDL and Verilog (one entity per IC, named after the circuit), simulated by Silinx as
    the engine. The MCP server: new_source type circuit, circuit_parts, edit_circuit, and circuits in
    check_drawing and simulate_drawing. The editor follows.
  • The circuit editor (#3898, part of #3897): web/js/circuit-editor.js and its modules (circuit-draw.js,
    circuit-toolbar.js, circuit-side.js): the breadboard, the ICs with their notch and pins, switches, buttons, LEDs and
    the supply drawn on the 0.1 inch grid; the parts library in the Properties panel, a part placed by a click, moved, turned
    (R), deleted; a wire drawn point by point, ended on a pin or a hole; Try it (each pin's level, the LEDs lit, the time in
    the status line, a button held while pressed, an oscillation or a short in the console); Check in the console; Undo /
    Redo of the main toolbar; the HDL file made on save. The Design pane's row [Circuit] … [VHDL] / [Verilog], New
    Source ▸ Circuit (in a design and simulation project too), its icon, Print, rename, the drawing kinds' lists of the
    project, the import, the design-only checks; its Portuguese labels (web/js/i18n-circuit.js).
  • The Arty S7 and the Zybo Z7 (#3895). core/boards-vivado.js: their variants, their pins from Digilent's master
    XDCs (the Basys 3's commit): the Arty S7's 12 MHz oscillator, 4 LEDs, 2 RGB LEDs, SW0..SW2 (SW3 and the 100 MHz
    oscillator are on bank 34 at 1.35 V: left out), 4 buttons, the USB-UART, Pmods JA..JD; the Zybo Z7's 125 MHz clock,
    4 LEDs, RGB LED LD6, 4 switches, 4 buttons, Pmods JA, JC, JD, JE (JB and LD5 are the Z7-20's only; its UART is the
    PS's). Their Pmods are headers (#3709): every example's ports get a pin, as Vivado's write_bitstream requires. The
    Arty S7's XDC configuration is its master XDC's (SPIx4 at 50 MHz); the Zybo Z7's has none (a Zynq's PL).
    test/vivado/examples.test.js builds the examples on them too (SILINX_VIVADO_BOARDS picks the boards): all ten routed,
    timed and written as bitstreams by Vivado ML Standard 2025.2, with no error; their known findings recorded
    (test/vivado-examples-config.mjs): ZPS7-1 on the Zybo Z7 (a Zynq's PL with no PS), vga's pace counter unused at the
    Arty S7's 12 MHz, blinky's debouncers dead at the Zybo Z7's 125 MHz (#3922). README's Vivado section.
  • Vivado ML Standard's 7-series (Part of #3895). The Spartan-7 a family of its own (FAMILY_INFO, the device
    database: its parts and packages from DS180, the Artix-7's HR SelectIO), and the Zynq-7000's XC7Z007S, XC7Z012S,
    XC7Z014S and XC7Z015 (DS190); the Spartan-7, the Kintex-7 (XC7K70T, XC7K160T) and the Zynq-7000 AMD Vivado's flow
    (server/vivado.js), experimental as the Artix-7, with no back end of the Silinx toolchain (no bit database of their
    parts): Synthesis Tool Vivado alone and locked in the dialogs, the project's synthesis tool ignored by Implement, the
    Silinx toolchain or ISE asked for refused with a message naming Vivado (POST implement, run_process,
    set_design_properties), none in the browser edition (its warning and its error). The Virtex-7, no free edition's,
    keeps ISE 14.7's flow. run.tcl of a Zynq-7000 sets none of an FPGA's configuration properties (CFGBVS,
    CONFIG_VOLTAGE, the flash's). list_devices marks the 7-series "experimental: AMD Vivado's flow".
  • A design and simulation project without HDL, part 5 (#3891): in a design project, Settings, Tools and the
    toolbar's gear have no Toolchain Settings; Tools, a module's menu and the Processes no (View) RTL Schematic; View and
    the Design pane call the Implementation view Design; Export Project exports a Silinx project, with no software to
    choose; the Files tab has no Language column; a symbol's Symbol Info no Xilinx library equivalents; Design Checks'
    message names Check Design. The MCP server's export_project makes a Silinx project there by default and refuses
    a Xilinx ISE one. Set as Top Module, the About dialog and the console's first line (the application's) stay.
  • A design and simulation project without HDL, part 3 (#3891): in a design project (the page's kind flag,
    web/js/project-kind-flag.js) the schematic's sheet leaves out its section "HDL kept with the schematic" and the HDL
    in its Description's note, an I/O marker its HDL type, a flip-flop's input its "no such Xilinx flip-flop", an
    instance's Generics are its Parameters; the palette has no HDL BLOCK, no Device Primitives category, a project
    module's tip its ports alone; More…'s and Curved wires' tips and Check's "HDL generation failed" leave out the HDL;
    the state diagram's description placeholder, the ASM chart's Generics (Parameters) and their hint, the truth table's
    tick boxes' tip; the state diagram's and the truth table's "Not in sync with x.vhd" is not shown. The MCP server's
    related_view results say Logic for HDL logic there and refuse the VHDL and Verilog notations, as the editor
    offers none.
  • A design and simulation project without HDL, part 4 (#3891): in a design project, a drawing's HDL made again
    says nothing (no "x.vhd generated / updated from …" in the console nor the status bar); one with errors warns
    "x.fsm.json: its module is not made again: …" (web/js/design-only.js hdlStatus, hdlWarning); Check Design's
    diagnostics (the console, the Errors and Warnings tabs, the simulation's elaboration errors) are on the drawing with
    no line, by Silinx (shownDiag), its process named Check Design, "Checking top module"; a drawing whose HDL file
    name is taken is [Not made], Make Module…; Find in Project leaves out the HDL; Commit lists no HDL file; Pull's
    merged drawing says nothing of its HDL; Save as Test Bench, Verify's counterexample and the stimuli editor's tip name
    the test bench, not its file; "a design and simulation project now" without its HDL. The MCP server matches: the
    drawings' sync messages (core/drawing-sync.js), check_syntax (Check Design, the drawing, Silinx), test_bench_wizard
    and verify (the test bench's name, no language), find_in_project without the HDL.
  • A design and simulation project without HDL, part 1 (#3891): the Symbol Info dialog (web/js/symbol-info.js)
    leaves out its Equivalent HDL section and the HDL in its parameters' note in a design project (the page's kind flag,
    web/js/project-kind-flag.js); the related views of a document (web/js/related-views.js, its tab's menu and View's
    Show … Beside) leave out a file of the HDL side, a translation and a C source's views. The MCP server already refused
    them there (read_file of a drawing's HDL, related_view translation).
  • A design and simulation project without HDL, part 2 (#3891): in a design project, the Schematic, FSM, ASM and
    Petri Net wizards' first page is "Name", with no Language field, their hints and the port table's without HDL
    (no "its HDL is generated from it", no std_logic / wire), their summaries and the Boolean Wizard's without the
    module's language (web/js/design-only.js langNote); New Project's note leaves out the HDL module; Design Properties
    says Top Module and has no Top-Level Source Type of HDL; Rename… of a drawn module asks its name alone (its file
    follows it, the console never names the HDL file, nor counts it); Source Properties… shows its drawing; Remove from
    Project names the drawing alone; the Design pane's badges' tips leave out the HDL; the Command Palette lists a drawn
    module's drawing, never its HDL file; "Project opened", the status bar and Open Project's list show no device (the
    project's kind instead). The MCP server matches: new_source refuses a language there, remove_from_project and
    rename_module name the drawing alone, list_projects gives the kind and no device.
  • New Source's File Name and Location page without its "Add to project" checkbox (#3890). The checked, disabled
    box is gone, and its Portuguese translation, which nothing else used. The MCP tool new_source had no such option
    (a new source is always in the project) and is unchanged. test/ui/new-source-no-add-box.test.js checks every type's
    page, in English and in Portuguese, and that the source made is in the project.
  • Regular expressions built from constants escaped completely (#3888, #4088, #4089): CodeQL's alerts 185, 186
    and 187 (js/incomplete-sanitization): test/classroom-template.test.js escaped only . in Silinx's version,
    test/mcp-schematic.test.js only ( and ) in its messages, and the Circuit editor's Portuguese labels
    (web/js/i18n-circuit.js) only (, ) and + in the parts' titles; each escapes every special character now,
    backslashes included. The titles' expression is the same (none has another special character). Tests:
    classroom-template, mcp-schematic, web-i18n.
  • The display's digit enables' level as a generic (#3884). The examples with a 7-segment display (traffic,
    blinky, pong, stopwatch, clock, isqrt, uart, micro80) have an AN_LOW generic (true: the Basys2's, a digit lit at
    '0'), which the top passes to its display (sevenseg's AN_LOW; blinky's and traffic's VHDL tops as an integer,
    1 or 0, to their Verilog display). A board's value comes from its an's
    activeHigh (core/board-generics.js), written as the top's default as CLK_HZ, BTN_LOW and SW_LOW are (#3710), and
    the console says "the board's clock, inputs and display". At its default, every example's netlist and bitstream
    are the ones before. Every board's segments and decimal point are active low: no such generic for them.
  • A state's and a transition's look in the FSM editor (#3883): the Properties of a selected state or transition
    end with Look (web/js/fsm-look.js): colour pickers (showing the colour drawn; ✕ back to the theme's), the line's
    width and style, the text's colour, size, style and font, and Default look; each change an undo step. The look is
    the object's look in the .fsm.json, only what differs from the default written (an old drawing drawn and saved as
    before); a colour not chosen follows the light or dark theme, a chosen one is kept. core/fsm-style.js checks every
    value (a colour #rrggbb, a size from 6 to 48, a width from 0.5 to 10, the lists) with the same messages in the
    editor (the console) and in the MCP's new set_fsm_look; a file's value not valid is an error of Check, the drawing
    made with the default look in its place. The drawing reads the look as CSS variables of each object (fsm-render.js,
    fsm-editor.css), the selection, the problems and the simulation still drawn over it; a circle grows with its text.
    Portuguese labels and messages. Tests: test/fsm-style.test.js (the values, the defaults left out, the round trip,
    the HDL the same in VHDL and Verilog, Check), test/fsm-i18n.test.js, test/mcp-fsm.test.js, UI fsm-look (each
    property drawn, refused, undone, the theme, saved and reopened, the HDL unchanged and compiled with no warnings,
    Portuguese) and fsm-side.
  • State diagram editor: a state's and a transition's outputs one per line; no Delete button in Properties (#3882).
    In the drawing a state's outputs are written one per line under its name (the circle grows to hold them), and a
    transition's label is its condition followed by " /", then its outputs one per line, the label growing away from its
    arrow; a state's Properties list its transitions the same way. The Delete state and Delete transition buttons of
    Properties are gone: a state or a transition is deleted as any selection is, by the Delete key, the toolbar's Delete
    or Edit ▸ Delete. Only the display changes: the files, the generated VHDL and Verilog and the MCP server's tools are
    as before.
  • The 7-series' .bit compressed (#3876): Generate Programming File with the Silinx toolchain writes the compressed
    bitstream, in the packets of Vivado's (multiple frame writes for the repeated frames), when the XDC sets
    BITSTREAM.GENERAL.COMPRESS TRUE, which the Basys 3's and the Nexys A7's constraints do; its log line says
    "compressed". The toolchain reads it back to the same frames (silinx-fpga's fragment).
  • nextpnr-himbaechel-xc7 0.3.0 (Fixes #3875). Its chip database times the routing, the slice's cells and a DSP48E1
    with no register by Vivado 2025.2's delays (#3516; the DSP48E1 was not timed at all, stopwatch's estimate twice
    Vivado's). Its placer has a delay estimate from the chip database's routes (#3803), and HeAP weighs a tile across like
    one up or down, with a refinement over 20 tiles (#3875). The stopwatch example places and routes with the best of 3
    seeds (impl.xc7Seeds, #3825): 101.21 MHz on the Basys 3, 100.84 on the Nexys A7 (seed 1 alone: 96.98 there).
    The timing's label says "conservative against Vivado" (it said "optimistic"), and so does the timing report.
    test/xc7-vivado-comparison.json's open flow re-recorded (its LUTs, frequency and compressed .bit, #3876; Vivado's
    columns the same netlists'); the test holds nextpnr's estimate between 0.65 and 1 times Vivado's on the same netlist,
    and every example at 100 MHz, with no exception left for stopwatch.
  • The 7-series' DSP48E1 timing, fitted to Vivado's (part of #3875). scripts/xc7-timing-fit.mjs's table for the
    nextpnr fork's chip database now has the DSP48E1's arcs with no register (its buses by name: A -> P 3,841 ps,
    B -> BCOUT 735 ps…, from Vivado's SDF of stopwatch, packages/silinx-fpga/research/xc7/timing-fit.json), the arcs the
    fork's xc7 0.1.2 left out: a path through a DSP48E1 was not timed, so stopwatch's estimate was twice Vivado's. The
    fork's silinx branch times the DSP48E1 by these arcs (its timing_fit.py, which the test holds the table to); the
    estimate changes once a release of it is pinned here.
  • The EDGE Spartan 6 board (#3873): a board entry from the vendor's master UCF (rev 3.2) and User Manual v3.1:
    XC6SLX9-2-TQG144, 50 MHz on P84, FT2232H USB JTAG and USB-UART, W25Q80DV SPI flash. Its LEDs share pins with the
    7-segment display and the LCD, so LEDs D2..D9 and the display are its resources and deviceNotes lists the
    shared pins. The clock, 7-segment, LED and button resources and openFPGALoader (cable ft2232: SRAM load and SPI
    flash) were checked on a real board.
  • The Spartan-3A Starter Kit's expansion connectors (#3872). The board's pin table (core/boards.js) has, after
    UG334 ch. 15 (figs. 15-11, table 15-1), its 6-pin accessory headers J18 (j18_io) and J20 (j20_io) and its FX2
    connector J17 (fx2_io, its 40 I/O; fx2_clk, its 3 clock pins), before J19 (through-hole pads, no connector), all
    as headers: #3709's rule puts the port bits the board has no pin for there, in this order. Every example's build
    for the board goes ahead, warning of each bit it put on a header pin (the display's 12, the buttons' and switches'
    the board lacks). The Silinx toolchain does not build for the Spartan-3A yet: on this board it is ISE's.
  • ISE's bitgen options by family (#3869). Silinx gave every Spartan-3 family the same bitgen options. The
    Spartan-3A's and 3A DSP's bitgen has no DCMShutdown (ERROR:Bitgen:38), and the Spartan-3's takes no
    ConfigRate:1 (from 3 MHz); the Zynq's has no DebugBitstream. They now get only what their bitgen takes: no
    DCMShutdown on the Spartan-3A and 3A DSP, ConfigRate:6 (its default) on the Spartan-3, no DebugBitstream on
    the Zynq. The Spartan-3E's options are unchanged. A test checks every option Silinx writes against each family's
    bitgen -help of ISE 14.7, and ISE builds the clock example for the Elbert V2 and the Spartan-3 Starter Board.
  • The Results document (#3864, part of #3191):
    • core/grade-class.js:
      • parseResults checks the results file: its shape and sizes. It is untrusted data.
      • classSummary sums it up: per test (Accepted, its rate, the most frequent other verdict, the weakest first) and
        per student. classCsv gives --csv's CSV, and classText the text.
    • web/js/grade-results-view.js:
      • the document: its toolbar's Export CSV, the two tables, its status line;
      • Class Results… in the Exercise document's ⋮ menu;
      • a project's grades.json or <name>.grades.json opened from the Files;
      • Portuguese labels.
    • The MCP server's grade results does the same from a file of the exchange folder, with the CSV (csv).
    • Tests: grade-class (five students graded for real, against the numbers counted by hand; the CSV; the file's
      problems), mcp-grade, UI exercise (both interfaces, Portuguese); docs/GRADING.md, docs/CLASSROOM.md.
  • Shared projects, phase 4c: the owner's room resumed on the relay (#3835, part of #3180):
    • The relay keeps the room for 2 minutes after its owner's connection is lost (server/collab-relay.js
      RELAY_LIMITS.grace). Its followers stay connected meanwhile, and are not told. Past that, they are told "the
      owner's connection was lost" (core/collab-room.js OWNER_LOST). Stop Sharing still ends the room at once. A room
      the owner's page hosts (#3227) has no grace: its owner is the page itself.
    • The owner resumes it (core/collab-client.js shareRoom's reconnect): the room's open is answered with a
      secret, and a new connection sends { type: 'resume', room, resume, client }. Its tries follow the followers':
      after 1, 2, 4, 8, then every 15 s. The room then counts each of its followers as a join again, so the owner sends
      each one the whole document: what changed meanwhile arrives. An editor's changes made meanwhile reach the owner
      (the editor's catch-up of phase 4a), and each follower's presence is sent again. A connection of the owner's that
      the relay has not seen lost yet is replaced by the resume.
    • The end: a room the relay no longer keeps (the grace past, the server restarted) ends the owner's session,
      said. The room also tells its owner why it closed, so that this is never taken for a loss.
    • Security: the keys and the signatures are unchanged, and the relay still reads nothing. The resume secret is
      128 random bits, given only to the owner (never in a link), and compared in constant time. A resume needs the
      server's token, as any open, and the room's client id. A wrong one is refused as a room no longer shared.
    • The protocol's version stays 4: the resume is a new first message, and the open's answer has a new field that
      older pages ignore.
    • The page (web/js/collab-share.js): "Reconnecting…" in the toolbar while lost (Portuguese: "A religar…"). The
      console says "WARNING: Sharing "": the connection to the relay was lost: reconnecting …" and "Sharing
      "": reconnected to the relay …".
    • The MCP (share_project): its share is resumed the same way. Its status says so meanwhile (connected: false, a line), and a change on disk meanwhile reaches the followers once it is back.
    • Tests: collab-room (the grace, the resume and its refusals, a connection replaced, the end past the grace),
      collab-resume (the owner's change meanwhile at the follower's, an editor's at the owner's, no grace, close() while
      lost, reconnect: false), collab (over the relay: a room resumed, the grace's end), mcp-collab (share_project cut
      and back), UI collab-reconnect (the owner's page cut and back in both interfaces, and in Portuguese).
  • Shared projects, phase 4b: a hosted room's follower found again at the meeting point (#3835, part of #3180):
    • When: a follower or editor of a room the owner's page hosts (#3227, WebRTC) whose owner was found at the meeting
      point loses its connection. Each new try looks for the owner there again, with no dialog, for up to 15 s
      (web/js/collab-meet.js findOwner's class's rejoin, followRoom's transport). Its tries follow the
      relay's: after 1, 2, 4, 8 and then every 15 s, for 5 minutes at most.
    • A transport that fails (the owner not found yet) is tried again later instead of ending the session
      (core/collab-client.js); one that gives no socket still ends it.
    • The codes exchanged by hand are not made again: they would have to be exchanged again. That session ends at the
      loss, as before.
    • Security: unchanged. Each rejoin checks the owner's proof (its signature of this follower's id and of their
      connection, with the link's owner's key), as at the first join. Every other peer at the meeting point is still
      ignored.
    • The MCP: follow_project does not follow a hosted room (it has no WebRTC), so it has nothing to make again.
    • Tests: collab-reconnect (a transport that fails, then joins; one that keeps failing, given up), collab-meet (a
      follower's connection lost, found again at the meeting point: the owner's change made meanwhile, each sees the
      other again), UI collab-meet (a follower's page cut and back: "Reconnecting to Ana…", the console's lines; both
      interfaces).
  • The import of an ISE schematic (#3828, core/isesch-nets.js): a module's block named with a range, reg(3:0), is
    an arrayed instance again (it became reg_3_0, one instance with width mismatches); its copies' names are claimed
    (its base may be a keyword, as reg in Verilog). A module's instance is drawn with its ISE symbol's pins on their
    sides, in their order, at their spacing (core/isesch-symbol.js iseSymbolLayout).
  • The import of an ISE project (#3828, core/xise.js iseModuleSymbols): a module's .sym drawn otherwise than
    Silinx's default symbol becomes its <module>.sym.json beside its source, so that its instances are not out of
    date; a file the import carries is kept (the server's and the browser edition's import).
  • The export to ISE (#3828, core/isesch-export.js): the names of a symbol's pins on the top and the bottom drawn
    inside its body (they were written beside the pin, outside it); a module's instances drawn with two symbols said (ISE
    draws them all with the first one's: Update Symbol first). The project's zip has no .sym.json (a Silinx-only file).
  • The 7-series' best of 3 seeds (#3825): an Implement Design property of the 7-series (impl.xc7Seeds: 3; Process
    Properties' "Place & Route (7-series):", in Portuguese too; the MCP's set_process_properties xc7_seeds): nextpnr
    (router2) places and routes the design with seeds 1, 2 and 3, the run with the best Fmax (nextpnr's estimate of the
    slowest clock) kept, a tie to the lowest seed, a failed seed left out (all failed: the first's error, as one seed's).
    The log says each seed's Fmax and the one kept, the timing report and the report's par.seeds too; only the kept run's
    warnings are given. The local server runs the three at once in worker threads (on a computer of 8 GB and 4 cores or
    more), the browser edition one after another in its worker; Stop stops them all and nothing is written. The browser's
    fallback to router1 (#3143) now says its warning among the run's lines, so the kept run's only.
  • The 7-series' examples against Vivado (#3760, part of #3145): scripts/xc7-vivado-compare.mjs builds every example
    on the Basys 3 and the Nexys A7 three ways: Silinx's open flow; AMD Vivado's own flow from the HDL, as the application's
    Vivado flow runs it; and Silinx's netlist placed, routed and written as a bitstream by Vivado (Vivado's DRC and
    write_bitstream's on Silinx's synthesis, its LUTs counted Vivado's way, #3160). The results are recorded in
    test/xc7-vivado-comparison.json, their table is in docs/TOOLCHAIN.md, and test/xc7-vivado-comparison.test.js checks
    them: Vivado's builds with the expected warnings, routed, met and DRC clean; Silinx's netlist through Vivado's DRC with
    its clock met; the open flow's LUTs and frequency within bounds of Vivado's. test/vivado/examples.test.js builds the
    examples on the Nexys A7 too. What it found: Vivado builds every example with only the known warnings on both boards,
    and accepts every netlist of Silinx's (no DRC error, its bitstream written); Silinx's LUTs are about Vivado's (1,308
    against 1,267 on the Basys 3), nextpnr's frequency estimate 1.2–1.75 times Vivado's timing of the same netlist, and
    stopwatch's DSP48E1s 2.2 times (#3875); Silinx's .bit is not compressed where the XDC asks for it (#3876). The
    comparison in the weekly Quality workflow is #3834.

Silinx is an ISE-style FPGA design environment that runs in the browser, made for teaching digital systems: VHDL/Verilog projects; schematics with live simulation; FSM and ASM state-machine editors, truth tables and Karnaugh maps, all kept in sync with their HDL; module and test bench wizards; behavioural and netlist simulation with waveforms; a board emulator; synthesis, place & route and bitstreams with its own toolchain (the Basys2 and Nexys2 boards) or with Xilinx ISE 14.7; and board programming, also from the browser.

Downloads

File What it is
Silinx.html Single-file edition, the same app as the site (https://pmnmalo.github.io/Silinx/). Download it and open it in a browser: design, simulation, synthesis, place & route and bitstream (the Silinx toolchain, Basys2 / Nexys2; its tools come from the site the first time). Projects are kept in the browser (File ▸ Download Project Bundle to back them up) or, in Chrome / Edge, in a folder on the disk. Programming the board's FPGA and Platform Flash over USB works from the site's address (Chrome / Edge, Chrome on Android).
silinx-<version>.zip Silinx run locally, with its dependencies: an alternative to the site (offline synthesis, projects as folders, the installed programmers, Xilinx ISE 14.7 modes). Needs Node.js 18 or newer. Unzip, then double-click Start Silinx.bat (Windows), Start Silinx.command (macOS) or start-silinx.sh (Linux) in the silinx folder; Silinx opens at http://127.0.0.1:8642. Projects are stored in ~/Silinx-projects.
silinx-docker-kit-<version>.zip Kit to build your own private Docker image of Xilinx ISE 14.7 (Windows, macOS Intel/Apple Silicon, Linux) for synthesis, place & route and bitstreams. ISE and its licence are not included and must not be redistributed: download them with your own AMD account. See ise/README.md.

The source code is attached by GitHub below (Source code zip / tar.gz).

Silinx is free software (GNU AGPL-3.0, Copyright 2026 Pedro Maló; see LICENSE, NOTICE and THIRD-PARTY-NOTICES.md). It is an independent project, not affiliated with or endorsed by AMD/Xilinx. Xilinx and ISE are trademarks of AMD.

Checking a download

The files' SHA-256 checksums (the file SHA256SUMS):

d2acdd2f99819f63dc135af137efb6614c508bcbf715cc3b780d78d5ae71ec4f  Silinx.html
e29fe976334e6a620a4c79a89d76511357d79f4c1591c31937dca99b18bd2772  silinx-15.26.0.zip
47a922a5a17e1e6f665ec5a296fa043d08bbdfced3755de7362584f9473720f0  silinx-docker-kit-15.26.0.zip

In the folder of the downloads: sha256sum -c SHA256SUMS --ignore-missing (Linux), shasum -a 256 -c SHA256SUMS --ignore-missing (macOS), Get-FileHash <file> (Windows PowerShell, compared by eye). Each file also has a build provenance attestation (it was built by this repository's Release workflow from this tag): gh attestation verify <file> --repo pmnmalo/Silinx.