Online Developer Meeting (2024-11-26)

From Octave
Jump to navigation Jump to search

Today's topics[edit]

  • Release of Octave 9.3.0?
    • Rik volunteered to shadow jwe while he makes release to learn the process. Markus and NickJ may also shadow event if the timing works out.
  • "Code sprint" for new terminal widget?
    • not discussed
  • buildbot: broken synchronization for stable branch?
    • Might have fixed itself by restarting the Digital Ocean server. (The server load was above 100 before the restart. Mainly Python and Apache processes again. Still suspecting the Mercurial web interface.) Will monitor the situation.
  • GitHub mirror:
    • broken synchronization for stable branch?
      • Fixed by Kai on Nov. 25th.
    • access permissions for organization members?
      • Base access permission for organization members has been changed from "write" to "read". Higher access permissions should be added on a per-repository level if required.
  • load path and PKG_ADD issues
    • Keep allowing arbitrary code execution in PKG_ADD scripts. Try to avoid the use-after-free when addpath is used in PKG_ADDscripts.
  • future of octave-cli (also: FLTK, and gnuplot graphics)
    • Look into moving FLTK and gnuplot graphics toolkits into Octave packages (if that is easy)
    • Eventually remove them from core Octave (regardless of whether they can be moved to packages).
    • Rename octave-cli to something that indicates that the main difference is that it doesn't link to the Qt libraries. Users of Octave with the CLI should *not* call octave-cli (but octave, potentially calling octave --no-gui) in the vast majority of cases.
    • After the "original" octave-cli has been renamed, potentially, add a new wrapper executable named octave-cli that essentially just forwards to octave --no-gui (to avoid potential issues with hard-coded executable names in build scripts or Makefiles).

Previous topics[edit]

  • Digital Ocean server is still slow intermittently.
    • At the time of the meeting, CPU load was 48 (100%)
    • Looking at the logs, the situation improved though.
    • Large number of Python processes (mainly for the web interface for hg.octave.org) are the main contributors to CPU load. Maybe, disable https access to hg.octave.org (only leave SSH)?
  • Octave 9.3.0:
    • Ready for release
    • jwe will start the release process shortly.
    • Could a person shadow jwe during that process to get an idea which steps are involved and could potentially to releases themselves?
  • Start process for Octave 10:
    • Merge default to stable (when?)
      • Maybe, delay the merge until the experimental command window widget is in a more usable state? Main blockers include: input blocks Octave indefinitely, displaying large output is very slow.
      • Tried to find "quick fix" for existing command window widget with limited (or no) success. Maybe, reach out to Microsoft developers at https://github.com/microsoft/terminal/issues/17054
    • Update gnulib
    • Pending changes?
  • To improve coverage for mkoctfile, should we include building some Octave packages in CI? (Maybe, the "stk" package because it builds .mex files, and the "control" package because it builds .oct files with C++ and Fortran?)
    • Also run tests for these packages in CI?
    • Fix version? Or always latest version from package index?

See also[edit]