On Sun, Jun 8, 2025 at 9:18 AM Damjan Jovanovic <dam...@apache.org> wrote:
> > > On Sat, Jun 7, 2025 at 2:01 PM Matthias Seidel <matthias.sei...@hamburg.de> > wrote: > >> Hi Damjan, >> >> Am 07.06.25 um 13:47 schrieb Damjan Jovanovic: >> > On Fri, Jun 6, 2025 at 8:32 PM Matthias Seidel < >> matthias.sei...@hamburg.de> >> > wrote: >> > >> >> Hi Damjan, >> >> >> >> Am 06.06.25 um 20:15 schrieb Damjan Jovanovic: >> >>> On Fri, Jun 6, 2025 at 7:55 PM Matthias Seidel < >> >> matthias.sei...@hamburg.de> >> >>> wrote: >> >>> >> >>>> Hi Arrigo, Damjan, >> >>>> >> >>>> Am 05.06.25 um 22:45 schrieb Arrigo Marchiori: >> >>>>> Hello All, >> >>>>> >> >>>>> On Thu, Jun 05, 2025 at 10:33:52PM +0200, Matthias Seidel wrote: >> >>>>> >> >>>>>> Hi Damjan, >> >>>>>> >> >>>>>> That sounds really good! >> >>>>> Yes! It makes the code easier to read, and getting rid of outdated >> >>>>> constructs is A Good Thing. >> >>>>> >> >>>>>> I would love to join, but I (still) don't understand enough of >> it... >> >>>>>> >> >>>>>> Would that be something to cherry-pick for AOO42X? >> >>>>> FWIW I think so. >> >>>> Done! (first 4 commits...) >> >>>> >> >>>> Unfortunately, I now get a crash when opening an ODP presentation on >> >>>> Windows... >> >>>> >> >>>> >> >>> Oh no! I'll investigate over the weekend. >> >> I get no error message and no crashdump... >> >> >> >> But it looks like it does not happen with every presentation. >> >> >> >> >> > Hi Matthias >> > >> > I cannot reproduce this on the latest trunk so far, but I will continue >> to >> > test. >> > >> > All .odp documents I found load without crashing, as does a new one I >> made >> > containing text and shapes. >> > One of my patches was to the linguistic module, but the api.i18n tests >> pass >> > and (the not yet pushed) api.linguistic2 tests pass. >> > bvt and fvt tests largely pass. >> > >> > My build flags were: >> > --with-dmake-url= >> > https://github.com/jimjag/dmake/archive/v4.13.1/dmake-4.13.1.tar.gz >> > --with-ant-home=/cygdrive/c/apache-ant-1.9.6 >> > --with-junit=/cygdrive/c/packages/junit-4.13.2.jar >> > --with-epm-url= >> > https://sourceforge.net/projects/oooextras.mirror/files/epm-3.7.tar.gz >> > --disable-directx >> > --with-jdk-home="/cygdrive/c/Program Files/Java/jdk1.7.0" >> > --enable-verbose >> > --enable-pdfimport >> > --enable-dbgutil >> > --enable-crashdump >> > --with-hamcrest-core=/cygdrive/c/packages/hamcrest-core-1.3.jar >> > --with-package-format=installed >> > --with-nasm-home="/cygdrive/c/Program Files/NASM" >> > --enable-category-b >> > --with-mozilla-build=/cygdrive/c/mozilla-build >> > >> > Your crash could be due to some specific difference, eg. fonts, >> languages >> > and locale, images, shapes, macros, etc. >> > >> > Please send me the documents that crash, if possible. >> > Share your build flags. >> >> I simply use the presentation wizard to create a presentation from >> template. >> >> After "create" AOO crashes immediately. >> >> I can only test on Windows and build with the configure I use for years: >> >> >> https://drive.google.com/file/d/1F8BlOJLbeE92ao6chdzJ6BUgCdP9iwng/view?usp=drive_link >> >> Regards, >> >> Matthias >> >> > I still cannot reproduce that Windows crash with trunk, but I have found > another way to crash it with the unit tests: > ./run > -Dopenoffice.home="../main/instsetoo_native/unxfbsdx/Apache_OpenOffice/installed/install/en-US/openoffice4" > -tp fvt.uno.sd > will crash in certain tests that didn't crash before. > > Reverting the following commit stops it from crashing: > > commit 1890034f73947872dc88f59af6226fed5174e2b5 > Author: Damjan Jovanovic <dam...@apache.org> > Date: Thu Jun 5 19:09:02 2025 +0200 > > Port main/linguistic component_getFactory() to the component context > API. > > Patch by: me > > I am investigating why. That commit seemed straightforward and unlikely to > cause problems, but maybe the extensive queryInterface() calls in > main/linguistic/source/gciterator.cxx behave differently with the new > service manager. > > > Reverting changes to gciterator.* and lngopt.cxx (and of course patching lngreg.cxx to call them the old way while still calling the others the new way) did not help. Reverting changes to gciterator.* and lngopt.cxx and lngreg.cxx, completely fixed the test crashes. Reverting only the changes to lngreg.cxx also completely fixed the test crashes, which tells us that the regression is in this simple harmless-looking part of the change: git show 1890034f73947872dc88f59af6226fed5174e2b5 main/linguistic/source/lngsvcmgr.cxx (and of course the related changes to main/linguistic/source/lngreg.cxx).
