Skip to content

Latest commit

 

History

History
189 lines (170 loc) · 8.65 KB

TODO.md

File metadata and controls

189 lines (170 loc) · 8.65 KB

TODO

Welcome to the braindump for current ideas in FVWM. These items are not ordered in temrs of their priority; rather they represent the key areas that FVWM could benefit from.

If you're thinking of working on any of these items, it might be worth dropping @ThomasAdam an email for discussion.

Items

  • Backend:

    • Wayland? -- maybe, but it's new and a moving target. No decorations (a la Xlib).
    • XCB
  • Parsing:

    • Print where an error was found whilst reading the config file (including line numbers).
      • What would modules do here? Think: *FvwmModuleAlias: foo bar
  • Code auditing/security:

    • Remove SAFEFREE() macro;
    • Add xasprintf() to libs/safemalloc.[ch]
    • Audit xmalloc() use and consider xasprintf() wrapper
  • Code cleanup:

    • Don't use typedefs for structs:
      • Opaqueness is not useful here!
        • 'struct foo' versus 'foo_t' or 'Foo' as the type.
        • Will allow for reducing the number of structs used as the main API, and will allow for better streamlining and accessing.
      • There's a lot of macro #defines for accessing struct members, presumably to hide the type; that's usually OK, but I'd like to see these go away. FW and Decor are good examples of this! I'd rather see more exposed functions to replace these. The API will need discussion. dv: Many of these macros were introduced to hide the layout from the underlying structures from its users. So, when the structures change - as they have many times - one would only have to change the macro definition and not hundreds of users. Especially with a future rewrite of the style system in mind, this is still very important.
    • What's in libs/ -- the static linking of libfvwm3.a useful anymore? There's a lot of code being ripped out of libs/ and I don't see much more being added in. dv: Static linkage was good when the library was small. Nowadays its just a burden.
    • I'm not keen on commands.h:P() macro, I'd rather have the prototypes explicitly stated.
    • Remove the ewmh_intern.h:CMD_EWMH_ARGS macro, I'd rather have the prototypes explicitly stated.
  • Messages:

    • More debug output to help the user pinning down why a window does not "behave". One problem that comes up all the time is "why does my window appear at position X Y with size WIDTHxHEIGHT". Although there is some output, when a window first pops up, there is none after the initial mapping when the really interesing things happen (messages from the client application).
  • String handling:

    • libs/String.[ch] and friends have idioms like CatString3, and other means to manipulating strings. Get rid of these.
    • Audit the code for strcpy()/strcat(), and take out back and shoot them! Specifically:
      • Check whether the original string is being malloc()d and just use xasprintf()
      • If it's on the stack, use strlcpy(), and carefully consider whether the return value should be checked. If it's not necessary, add (void) at the start of the call.
    • Each command shouldn't be doing its own string parsing; the items should be tokenised before-hand and sent along to the relevant CMD_*() functions.
      • This means rethinking how commands/lines/etc., are parsed.
  • RandR:

  • Clients:

    • Have an "undo" list of geometries which can be reverted to.
    • Per-page geometries for windows, makes handling sticky windows easier.
    • No special-casing of maximised state, it's just another geometry set which is added on to a list, and popped when no longer used in that state. dv: Unfortunately its not that easy because being "maximised", "shaded", "iconified" or "fullscreen" is announced to the world through the EWMH interface. The list approach sounds promising, but one has to keep in mind that the geometries are not independent of each other. For example, a window is halfway between two pages, maximised on the current page, then the user switches to the other page and unmaximises the window. Now, where dos he expect the window to appear?
  • Commands:

    • There's far too many commands in FVWM:
      • Too many duplicates; merge.
      • Think about a common way of referring to pages/windows/etc -- much like with tmux's "-t session:window.pane" syntax, can we not do something similar here for referring to pages/windows?
    • Handling of command options:
      • Commands don't do their own parsing; instead, command tokens are parsed with getopt(), and a tree of key/values are built; that tree is passed in, and referenced arg_has()/arg_get(), etc.
  • Functions / Conditionals / Exec (PipeRead) / Etc.: * Next/Prev/etc., are all very scripting-like. Do we just want hook support? Is it worth expanding FvwmEvent to do this?

  • Menus:

    • Typeahead/autocomplete inside of menus. See 'rofi' for example of functionality.
    • Potentially consider appmenu for a global menu for the focused window.
  • Modules:

    • The module interface (FVWM <-> Module) is a mess; consider:
    • DBus
    • imsg
    • Msgpack-C: (https://github.com/msgpack/msgpack-c/blob/master/QUICKSTART-C.md)
    • Cap'n Proto (bloody stupid name!): https://capnproto.org
    • Use libevent to replace the hand-rolled (and often broken) select/poll mechanism.
    • What about third-party scripting languages? How do we handle that without requiring linking against the specific language in question?
    • Amalgamation of module(s) is a possibility:
    • Functionality for things like buttons, taskbars, iconman, pager, etc., can be replaced with a single module:
      • Combines all the functionality of the obsolete modules.
      • Uses colorsets consistently, to allow configuring all GUI elements.
      • Can be reconfigured at runtime through user interactions (e.g. dragging elements to a different position on the window)
      • Initially support the older module interface.
      • Can replace all the old modules with symlinks.
  • perllib (FvwmPerl):

    • Transition away from 5.004 as the minimum version;
    • Update the code to use 5.14 at least:
      • Do away with subroutine prototypes;
      • "use vars" is deprecated;
      • "use parent", rather than manipulating @ISA/Exporter directly;
      • "use warnings/strictures" everywhere;
    • Don't rely on the crappy command generation stuff from FVWM releases to feed the perl API ""abstraction"" as we have it now (deprecate create-constants and create-commands).
  • Decors: dv: I think decorations should be dropped in favour of styles. The whole concept was invented because stlyes were unable to do it.

    • Think carefully about the syntax of the decor replacement!
    • Remove {Border,Title,Button}Style and decors, just one simple look.
    • New style DecoratedByFvwm/DecoratedByModule.
    • New module FvwmDecor to do per-window decorations:
      • How does resizing work with this module?
  • Styles: dv: I have much more radical plans for styles: Being able to nest styles, making them conditional with any condition that can be used anywhere in fvwm at any time, defining window specific preferences of style order.

    • Expand Style command to support name=pattern syntax, for instance: Style (Name "File *", Class XPaint, Resource *browser) NoIcon
    • Add id-pattern to Style command.
    • Split styles into five lists (resource, class, icon, name, id) which are then applied in that order.
    • Split Style into run time control (Style) and startup style (InitialStyle).
    • Remove the ability to toggle specific parts of EWMH; it's either on or it isn't. dv: I don't think this should be done. EWMH is annoying in many ways, and in order to tame EWMH "enabled" applications, detailed control over the various aspects. It was a single switch originally, and that was not good enough.
  • Add missing translations:

    • StaysOnTop
    • StaysPut
    • StaysOnBottom
    • Identify