Developer Meeting Agenda 2026-11-08 - 10:00 CET

[grandixximo]

According to our bi-weekly schedule alternating between "early" and "late", this video meeting is Sunday 2026-Nov-08 at 10:00 CET.

https://greenlight.bbb.uni-rostock.de/b/ste-c4d-brs-3k6
Access code: 869782

https://www.timeanddate.com/worldclock/meetingdetails.html?year=2026&month=11&day=8&hour=9&min=0&sec=0&p1=319&p2=236&p3=240&p4=136&p5=165&p6=256

Agenda

0. Conclusions of 2026-10-25: nothing was posted in Developer Meeting Agenda 2026-10-25 - 20:00 CET #4705. Could someone who attended post them there, so the items below start from the right state?

1. Plan release 2.10, carried from 2026-10-11 item 13 and 2026-10-25. Decide: branch now, freeze master first, or branch once a named list is done, and which list? Since last time the HAL 64-bit component series from @BsAtHome landed (hal: Update hm2_modbus driver status params and pins to use 64-bit #4740, hal: Update streamer/sampler to use 64-bit accesses #4742 to hal: Update to 64-bit clean operation and accesses #4745, hal: Update debounce and modmath to run 64-bit clean #4772, hal: Update the pid component to 64-bit clean operation #4773; hal: Update sim_encoder to 64-bit clean operation #4771 sim_encoder still open).

2. NML replacement and the UIs built on it, discussion NML replacement #4571.

   • Out-of-tree UIs are appearing now. @rmu75 pointed at LiNC (Rust and egui, started 2026-10-22), which talks to LinuxCNC through linuxcnc-nml, Rust bindings over libnml covering the whole emcmodule.cc surface (all linuxcnc.command methods, all linuxcnc.stat fields, the error queue). On the forum there are also web UIs (lcncgui, lcnc-suite) on top of the Python module. Each of them breaks when NML goes.
   • Do we commit to keeping the client surface stable across the replacement: the Python linuxcnc module as it is, and maybe a C ABI client library with the same commands and status, so bindings like linuxcnc-nml can move over instead of being rewritten?
   • Requirements: @rene-dev wrote on 2026-10-28 that he has a good understanding of what is needed; @BsAtHome asked for it in detail. Who writes it down, and where (NML replacement #4571 or a new issue)?
   • Related open work: Zmq halui gui #4613 (ZMQ halui GUI), WIP: halui: pin-only GUI bridge surface (halui side of the #4613 split) #4616, Gui panel pins #4653 (GUI panel pins), modeless task, and jog-while-paused #4700 (modeless task, jog while paused), Unify the way jogging works from HAL, and its interaction with the UI #4692 (jogging from HAL), Unify the way MDI commands are executed from HAL / GUI #4691 (MDI from HAL and GUI).
   • 2.10 and out-of-tree clients: since build: Fix include copying, references and RT/non-RT isolation #3837 the NML headers are not exported, so linuxcnc-nml cannot build against master (Some development headers missing from make install #4774). build: export the NML client headers again #4785 installs 21 of them for as long as NML exists, with no layer on top and no emulation promised (@rene-dev, @rmu75). @BsAtHome cannot approve that: exported headers must move into include/ and be cleaned of internals first, and Python is the published client API. Decide: a time-boxed export, the structural move first, or Python only for 2.10, and does it block branching?
   • Architecture: which state and which inputs belong to the core and which to each UI (sketch). Agree on this, or a better split, before the requirements in NML replacement #4571 are settled.

   ---

   Conclusion: Export the headers for 2.10 but split, @basathome will do the split, 3.0 new API for UIs, websocket 2.10 maybe
   

3. Buildbot (buildbot2.highlab.com).

   • There are no Debian 13 trixie builders: the newest are bookworm (10-rip.debian-12-bookworm-rtpreempt-amd64, 50-deb-uspace.debian-12-bookworm-amd64, 30-dsc-uspace.debian-12-bookworm). GitHub CI builds trixie packages (package-arch, package-indep, 32-bit, rip-rtai), but runs the tests only on Ubuntu runners without a realtime kernel, and the buildbot is where the tests run under SCHED_FIFO. Can we get a trixie rtpreempt worker (rip with tests, deb, dsc)? Who has admin access to the buildbot and its hosts?
   • Retire the wheezy, jessie, stretch, buster and precise builders?
   • Two builders have failed in every one of their last 60 builds, back to at least 2026-07-22:
     • 50-deb-uspace.debian-12-bookworm-amd64: lintian errors executable-in-usr-lib on the qtvcp Python files (qtvcp/lib/qt_vismach/*.py, ripper/gcode_ripper.py, writer/writer.py).
     • 20-docs: the docs build, but the rsync to www.linuxcnc.org fails with Permission denied (publickey,password). Is the docs site still being updated, and whose key is it?

---

Conclusion: Seb has control, we cannot do anything, Rene would like to put a debian 13 machine hooked to github CI

4. The wiki, discussion The wiki situation #4767 (carried from 2026-10-25): https is broken and nobody has admin access at dreamhost. Retire, move (dokuwiki was suggested), or fold what is still used into the docs? @andypugh named pages that only exist there (ContributedComponents, MinimumSoftwareVersions).

---

Conclusion: Try to move it to github Linuxcnc organization Pages. Cleanup and remove old stuff, and update, Rene and Robert are on it

5. Forum (carried from 2026-10-25, mailing list thread "Critical PHP Vulnerabilities"): the forum is updated, still on Kunena. phpBB was floated on IRC; Discourse looks too heavy for the server. Stay on Kunena, or move, and who would do it?

---

conclusion vulerability is dealt with we stay on Kunena for now

6. Not reached on 2026-10-11, still open:

   • Development model, discussion Branch Protection #4558: areas and maintainers, what goes through a PR, cross-area changes, CI and branch protection, where it is written down.
   • Versioning and release procedure, CI build artifacts on master have wrong version #4555 and Rework versioning scripts and workflow #4584.
   • HAL names, Feature request: (Not)Accepting decoration as a hal signal name when there is no space #4645 (decoration at the end of a name) and hal - alphanumeric sort #4650 (natural sort in the tools).
   • mc_axis proposal, discussion Proposal: mc_axis — a PLCopen MC Part 1 axis component for HAL (extra joints, axes without G-code) #4647.
   • Logging library, discussion Logging library #4638.
   • Signed vs unsigned integers in HAL, The unsigned integer question #4526: who takes the conv_* components, the cross-net warning and the docs?
   • Meeting day and time (@rmu75 on the list).

7. HAL 64-bit, remaining:

   • Draft - HAL types update - Remove 32-bit and move to 64-bit #4247: close it, now that the break and the component series are in?
   • Converters hal: Add halfileupdate to convert HAL configuration files #4499 (halfileupdate) and hal: Add halcompupdate to migrate .comp files to the new HAL API #4256 (halcompupdate) are merged. update_ini calling halfileupdate for the 1.1 to 1.2 upgrade: asked on Unify INI values #3704, waiting for an answer.
   • hal: C++ API (hal.hh) and pybind11 bindings on the new HAL API #4251 C++ API and pybind11 bindings, needed before branching 2.10 per 2026-10-11 item 4. Rebased on 2026-11-05, waiting for review.

   ---

   hal: C++ API (hal.hh) and pybind11 bindings on the new HAL API #4251 will hold on Bertho's review

8. [DISPLAY]CYCLE_TIME, GUIs: read [DISPLAY]CYCLE_TIME with one rule #4677 (item 10 of 2026-10-11, decided on 2026-09-13). The PR implements the rule for every GUI and VCP, and following review writes seconds everywhere (73 sample configs move from 100 to 0.100). Behaviour changes to agree on:

   • The default becomes 0.150 s everywhere, so configs without CYCLE_TIME change: AXIS was 20 ms, QtVCP, GladeVCP and gscreen 100 ms, tklinuxcnc 200 ms.
   • Invalid values (zero, negative, below 0.001, non-numeric) now warn and fall back to the default.

9. PRs that change behaviour and need a decision:

   • Priority: interp: G43.1 and G43.2 axis words are offsets, never a move #4675 G43.1 and G43.2 axis words are offsets, never a move (stray motion, hit on a real machine by Sigma1912). @andypugh: would you like a slimmer 2.9 backport with only the G43.2 fix?

   ---

   Will backport to 2.9 so that there it does not move as well

   • interp, motion: the point-to-point moves and the tool orientation codes, G53.1 to G53.7 and G68.3 #4730 point-to-point moves and tool orientation codes, G53.1 to G53.7 and G68.3 (item 5 of the multiaxis roadmap Multiaxis kinematics and limits #4374). Large, no review yet: how do we review it?
     --- check conflicts
   • motion: keep the last pose as the forward's guess after a failed forward #4757 motion: keep the last pose as the forward's guess after a failed forward (all KINEMATICS_BOTH machines).
     --- holding for review, problem is real
   • homing: run every leg of a synchronized home sequence together #4577 homing: run every leg of a synchronized home sequence together.
   • xhc-hb04: rename dashed ini identifiers, migrate old configs in update_ini #4601 xhc-hb04: rename the dashed ini identifiers and migrate old configs in update_ini (follows the strict identifier decision).

   ---

   ready to flatten TCL

   • preview: one preamble for every GUI, restating modes only #4618 preview: one preamble for every GUI, restating modes only. Needs a look from the GUI side.
   • 2.9: interp: RETAIN_G43 keeps the offset of the loaded tool #4787 (RETAIN_G43 lost the tool length after a real tool change, and failed G41/G42 otherwise; changes behaviour), gremlin: do not restate G41, G42 and G96 in the preview init code #4788 (preview failed at line 0 under G96, G41 or G42), interp: an abort under G96 restores the other modes again #4789 (fixes the abort restore lost under G96, a regression from my 2104c8b), task: undo read-ahead modes when an error aborts a program (#4715) #4727 (undo read-ahead modes when an error aborts a program, RESET STATE IF NOT USE IN REMAP #4715), configure: fall back to the default scheme for SITEPY (2.9 branch) #4790 (SITEPY on Gentoo, Gentoo bug: ModuleNotFoundError: No module named 'linuxcnc' #4717), motion: drive joint.N.index-enable only on edges (backport #4200 to 2.9) #4204 (backport of motion: drive joint.N.index-enable only on edges (fix #3556) #4200, index-enable on edges), xhc-whb04b-6 fixes backport to 2.9 #4761 (xhc-whb04b-6 fixes backport, approved). Merge branch '2.9' #4759 merges 2.9 into master and waits for xhc-whb04b-6 fixes backport to 2.9 #4761. Is another 2.9 release planned?

10. PRs waiting for a reviewer or an answer:

    • Canon cleanup #4748 canon cleanup.
    • xdr upstream readable #4762 XDR configure for macOS and other platforms.

    ---

    holding on re-review @BsAtHome

    • hostmot2: fix PktUART queued counts and byte packing #4659, hostmot2: bound Ethernet packets and check transfers #4660, hostmot2: harden Smart Serial and absolute encoder handling #4662 hostmot2 PktUART, Ethernet and Smart Serial hardening. (still being reviewed)

    ---

    holding on review

    • motion: check circular moves against axis soft limits #4619 circular moves against soft limits (on hold, @rene-dev has a simpler idea).

    ---

    holding for @rene-dev shorter version

    • tp: allow tangent blending across near-collinear ABC and UVW motion #4421 tangent blending across near-collinear ABC and UVW motion.
    • Add Support for Dogbone Homing #4009 dogbone homing.
    • Add generic and/or with flexibel inputs & allow unconnected pins #4746 and_n / or_n.
    • Realtime config scripts & doc #4752 realtime config scripts (draft).
    • linuxcnc: default GDK_GL to software-draw-surface #4777 default GDK_GL to software-draw-surface.
    • Show the estop reset after a reset and machine on between two status polls #4778 show the estop reset after a reset and machine on between two status polls.

    ---

    will look at this again, if it can be fixed only in gstat

    ---

    Side note Interpreter parses _hal and it should be done differently. @rene-dev will look at it

[Sigma1912]

I know, I repeat myself but given that this bug causes stray motion it would be nice if this fix could be given some priority:
#4675
Thank you.