Description: Test build of libreoffice 5.4.2 in Ubuntu Bionic (18.04) against libicu 60.1: Test name: TestBreakIterator::testLineBreaking equality assertion failed - Expected: 4 - Actual : 9 - Expected a break at the first slash This is a regression, the test was passing when building against libicu 59.1. Steps to Reproduce: See full build log at https://launchpadlibrarian.net/344849432/buildlog_ubuntu-bionic-amd64.libreoffice_1%3A5.4.2-0ubuntu3_BUILDING.txt.gz Actual Results: Line break doesn't happen where expected. Expected Results: Line break should happen at the right place. Reproducible: Always User Profile Reset: No Additional Info: User-Agent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Ubuntu Chromium/63.0.3239.30 Chrome/63.0.3239.30 Safari/537.36
Hello Olivier Tilloy, Thanks for reporting this issue. Since this is not an issue in LibreOffice's software, I'd recommend you to send an email to the dev mailing list [1] instead, the right place for this kind of problems. Closing as RESOLVED NOTABUG
09:56 < ricotz> hello, how is this not an upstream problem https://bugs.documentfoundation.org/show_bug.cgi?id=113705 ? of course this is building with icu 60.1 which isn't officially used by upstream yet 09:56 < IZBot> bug 113705: LibreOffice-framework normal/medium RESOLVED NOTABUG linebreak unit test failure when building against libicu 60.1 09:57 < ricotz> _rene_, hi, fyi ^ 09:57 < tml_> but the resolution says it *is* an upstream problem 09:58 < tml_> in icu upstream, that is 09:58 < loircbot> LibreOffice (core) caolanm * vcl/unx/generic/gdi/cairotextrender.cxx: ofz speed up fuzzers a tad 09:59 < _rene_> tml_: why should it? 09:59 < ricotz> tml_, I assumed this is some behaviour change which libreoffice would need to adapt to 09:59 < _rene_> exactly 09:59 < tml_> or is there something sacred with the ICU software that they can never have bugs, and any regressions in their behavioiur is always the fault of their clients? 09:59 -!- ohallot [~Thunderbi@179.218.180.46] has joined #libreoffice-dev 09:59 -!- iplaw67 [~Thunderbi@178.73.157.19] has quit [Read error: Connection reset by peer] 10:00 < ricotz> tml_, this has to be determined without closing this bug prematurely imo 10:00 < _rene_> tml_: you don't know. but the NOTABUG based on x1sc0 s comment is simply wrong 10:00 < _rene_> it mnight be a ICU change LO needs to adapt to, yes 10:00 < _rene_> like the changes in 59 etc
10:07 < khaledhosny[m]> There are line breaking and other text segmentation changes in latest version(s) of Unicode that ICU probably implemented and our code mught having wrong assumptions about. assign to Eike
Those are the changes related to line breaking in unicode 10: http://www.unicode.org/reports/tr14/tr14-39.html#Modifications
What's the priority for that issue? Can we expect an upstream fix in the near future (if so which timeframe), or should I try and come up with a distro-patch for Ubuntu (I'm not familiar with that code, I don't how much work that might involve)?
Might not be a change in line breaking rules after all. ICU changed something related to break iterators and that might be the root issue here, or it might be just a pure ICU bug. Needs someone familiar with the use if break iterators in LIbreOffice to debug and figure if it is a bug in ICU or in our code.
There's https://gerrit.libreoffice.org/#/c/44665 that updates the bundled version of ICU to 60.1, but it fails to build even earlier.
https://cgit.freedesktop.org/libreoffice/core/commit/?id=9206a08ada00e8762c4a634f242bd566028964bb