COOL TC meeting minutes: 2026-08-12

Development
1

Date: 2026-08-12

Present:

  • Regina, Stephan, Szymon, Andras, Michael S, Thorsten, Noel, Miklos

Completed action items

Pending action items

Incremental builds in CI (Stephan)

  • Incremental build is rather useless, because it “false positives”
    when there is a problem, but an incremental build doesn’t find all problems

  • Had several cases in the past where this caused trouble

  • We have 10 checks for the ‘main’ branch on gerrit (Andras)

    • Know that Windows is from-scratch now

    • Need to check if the rest are also non-incremental

    • AI: check that no builds are incremental anymore (Andras)

clang loplugin in code outside engine/ (Miklos)

  • enabled when engine/ is build in dbgutil mode with clang, but no CI for this?

  • Simply the engine/ loplugin was not working outside engine/ when the CI job was set up (Andras)

  • **AI: **will look at this (Andras)

clang-format (Stephan)

  • there’s a git commit hook, but people seem to manage to get past it (and our CI has no git-format check)

  • so whenever I want to commit something with the hook intact, it breaks and reformat-formatted-files keeps reformatting unrelated things, which is a pain …

    • should we get rid of the clang format thing or – have a CI check that clang-format checks ?

    • The same as the hook would check → rejected by CI.

  • Clang-format doesn’t have a stable output (Miklos)

    • which makes it hard to adopt new C++ features etc.

    • add C++ ‘concept’ it gets mis-formatted – and we can’t upgrade
      the tool to the fixed version

    • would propose disabling it.

  • Tried it for a decade → lets admit defeat and move on (Thorsten)

  • covertly moving us towards declaring defeat → eventually exclude list contains everything (Stephan)

  • Decision: disable clang-format – but encourage using it once for entirely new modules.

  • **AI: **do the removal (Stephan)

    • leave the config file …

com.sun.star vs cpo (Noel)

  • Incremental transition is a bit painful.

  • Stopping incremental builds stop random build breakage we hope (Michael)

    • also something with PCH headers (Stephan)

      • seeing those often in recent times – why it fails – cpo renaming related.

  • Does CI do things right ? (Michael)

    • Windows CI doesn’t bother me (Noel)

      • Windows CI now takes longer – bottleneck is Windows CI …

        • 4-6 hours … we kept up previously with this.

        • Now Windows takes much longer and is the limiting factor – 40 mins each.

          • It was 20-30 before with incremental builds.

          • It catches up – morning queue was empty again.

      • Can dedicated some of the new H/W to this perhaps (Andras)

  • cpo renaming – compatibility with basic & python scripts (Noel)

    • for the old names … (Thorsten)

    • incremental bits → we have a confusing picture

      • if happy to break customers code – do it in one step.

      • Changed all com.sun.star → cpo

      • incrementally – hard spot to tell people.

    • Communication things work … (Stephan)

      • if add something – old names still resolve – zero point doing the renaming at all.

      • Was always against it

  • Can we add a simple bridge for this ? (Michael)

    • how safe is this practically ? (Stephan)

      • putting work on our shoulders … no reason to put on without the rename.

  • Strong preference to stay compatible while incrementally moving (Michael)

    • by switching a single flag → from version XYZ on – we change this (Thorsten)

  • How many entry points are there to tweak ? (Michael)

    • for each of the scripting bindings – something for basic / python – and for new JS stuff that does this transparently from CSS → CPO where it is needed.

    • And for where clients actually compare strings – never guaranteed to work.

      • Not sure how wide that scope is there.

      • Just a string in one place - can or cannot change the IDL …

      • no urgency in renaming the impl. names perhaps

        • not been touched yet – just types (Noel)

        • have massive typos in, someone spell-checked them

        • Whomever compares impl. names – has issues anywhere (Michael)

  • **Decision: **try to do compatibility pieces for Basic, Python, JS first, and
    then move incrementally as before.

Release Engineering update (Andras)

  • Releases

    • COOL 26.04 release red code from the main branch: tagged cp-26.04.3-1.

    • co-26.04 branch was reset to main on Friday

      • share/demos – is red-code which is on main …

      • We do need to back-port stable fixes to 26.04 from here … also for CODA & Mobile.

Patch review (Miklos)

  • All changes

    • Week 33: 187(-66)

  • PRs to migrate from GitHub

    • Week 33: 4(+0)

  • New contributors since last week

    • Week 33: HammadMengal, Tinde (Joke)

  • Top 10 reviewers since last week

    • Week 33: Miklos Vajna (92), Noel Grandin (49), Tomaž Vajngerl (42), Mike Kaganski (19), Michael Meeks (18), Michael Stahl (16), Stephan Bergmann (16), Szymon Kłos (8), Andras Timar (4), Henry Castro (3)

Bug reporting

Jenkins / CI update (Adria)

Crash testing (Caolan)

  • 2026-07-29: 62 (-6, +13)

    • backsliding with apparently old layout issue surfacing after numbering improvements

  • 2026-07-22: 55 (-37, +0)

    • more good progress

TTT talks

Binary file formats & autosave

  • We load file formats with very bad and obsolete password protection.

  • Should we load legacy file formats as read-only and force people to save as OOXML when upgrading binary filters.

    • Enabling us to remove the binary export filters.

    • The binary formats don’t have new features since Office 2003 or so …

    • Lots of things the users might insert can’t be stored in those formats either.

  • **Decisions: **have a dialog recommending ODF or OOXML for save … and loading readonly

What’s cooking