Status report: consultancy team, week 46, 2023

Status reports
1

Caolan

  • try and finish cypress optimization regression tests
  • check latest crashtesting failures
    → ItemSet rework has triggered an assert
  • have a look at apparent richdocumentscode bug
    • plausible fix looks fairly easy
    • trying to quick and easy test it though is a big of a
      struggle
  • tdf#158121 fix a regression pinned on me
  • unexplained updates are making it unreliable to
    add a regression test for save + multi user
    join + editing to ensure we don’t get full
    document invalidations.
  • oddly doesn’t happen in a manual session for me
  • but does in cypress
  • debug
    → Its the grammar checker, and cypress is en-US
    while my session is a different language without
    grammar checking
    → turn off spell/grammar checking for the tests,
    should work on reducing/eliminating those invalidations
    if there is nothing that requires it. But not right now.
    → get all that merged
  • DeepL stopped working, seemed to be fixed in the
    desktop version, but online still not
    → needs certs, can copy the first suitable candidate that
    exists that will work into the jail from system.
  • bring 3 formula updates per keystroke to 1
    → https://gerrit.libreoffice.org/c/core/+/159424
  • pull wasm toolchain builder images, etc.
  • build online wasm
  • try using the build wasm stuff to figure out how if it works at all
    → huh, it seems to actually work.
  • try and get nextcloud working with
    WASM headers for online · Issue #3258 · nextcloud/richdocuments · GitHub applied
    → doesn’t apply cleanly, simple mod to get it to apply doesn’t run
    no obvious explanation in logs
    → but requires https://github.com/nextcloud/server/pull/38082 so
    that’s probably why, hard to backport all this to current
    nextcloud release. Turning into a mess
    → better off in retrospect to use a current snapshot and start with
    that
  • spend some time interactively testing calc, a lot of weirdness
    getting reported recently. Do we have a problem when a user goes
    idle, either then or when a session restores?
  • doubt cast that the wasm stuff is actually working the way it seems
    to work for me. But it certainly seems to be executing fine with
    the server shutdown and there are very wasm-y looking messages in
    the browser console.

Mike

  • Question: After saving signed document, signing is
    removed, but macros remain signed
    • Asked Miklos to review the chain => merged
  • Libre Office build without ssl implementations
    • Merged the last patch (for tests)
    • Finished
  • tdf#157326 - so excited to finally have a reliable reproducer
  • Exporting multiple consecutive spaces to ReqIF-XHTML
    • Similar to my old https://gerrit.libreoffice.org/c/core/+/41444 :smiley:
    • I worry if the embedded elements (like
      and objects) can create
      newlines in paragraphs, which would break paragraphs with
      xml:space=“preserve”…
      • => easy for
        , but problematic for objects

Tomaz

First-page header and footer from and to DOCX

  • Moved more tests that failed to header/footer test file - now 6 tests fail
    but they duplicates (fail for the same reason)
  • Refactored to use struct of bools instead of binary flags for setting
    left, right, mirroring of pages and sharing (header, footer, first) so it is
    easier to debug
  • Fixing the test that has multiple page styles, which are copied and the
    last page style is not correct
    • we set sharing “correctly” (probably)
    • it is very confusing that the XText objects from properties
      “HeaderText”, “HeaderTextLeft”, “HeaderTextFirst” depend on what the sharing
      properties are set
      • doesn’t help that sometimes we need to set the a property even when it
        is actually disabled (“HeaderTextLeft”, but the “HeaderIsShared” is set to
        true).
      • need to bend backwards to get this to work
    • test is correct if the order of setting the “HeaderText” and
      “HeaderTextLeft” is reversed
      • annoying the order is important - it shouldn’t be
    • need to go deeper into the Writer code to see what is going on here to
      resolve this scenario
    • happens with the “Nightmare” use case

Miklos