Kornel Benko wrote: > Am Sonntag, 24. April 2016 um 11:20:16, schrieb Georg Baum > <georg.b...@post.rwth-aachen.de> >> Kornel Benko wrote: >> >> > Am Samstag, 23. April 2016 um 12:01:32, schrieb Kornel Benko >> >> Hm, there are more differences. In cmake we add also header files to >> >> the list. But I wonder, why unix command 'sort' is not working the >> >> same as cmake internal sort. >> > >> > OK, setting env LC_ALL=C, sort behaves under linux as the internal sort >> > in cmake. >> >> Not for me. Note that we do already set LC_ALL=C internally before >> calling sort. I also tried to set LANG and LANGUGAGE in addition, but >> that does not change the behaviour. I still get the same sorting for >> fr.po as in 79074e6c with autotools, and with cmake I get the sorting as >> in a32113e6d. >> > > Hm, the differences I get is not because of sort, but because under cmake > we check many more files.
Not in my case. > Also some filenames in source/po/POTFILES.in are wrong (because not > existent) E.g. > src/frontends/qt4/GuiPrint.cpp > src/rowpainter.cpp > > I don't know what else could be the culprit Did you run autogen.sh after the last newly added or renamed file? It is needed to generate POTFILES.in. Do you have any file in your source tree that is not in git? That would unfortunately be picked up by autogen.sh as well. > Comparing the actual (2.3-staging) fr.po with created one in cmake, I see: > diff fr.po /usr/src/lyx/lyx-2.0.x-git/po/fr.po > ... > 355,361c355,363 > < #: src/frontends/qt4/ui/InsetParamsUi.ui:138 > src/frontends/qt4/ui/LogUi.ui:88 < #: > src/frontends/qt4/ui/ParagraphUi.ui:261 src/frontends/qt4/ui/PrefsUi.ui:90 > < #: src/frontends/qt4/ui/RefUi.ui:343 > src/frontends/qt4/ui/SearchUi.ui:203 < #: > src/frontends/qt4/ui/ShowFileUi.ui:55 src/frontends/qt4/ui/SymbolsUi.ui:66 > < #: src/frontends/qt4/ui/TexinfoUi.ui:58 < #: > src/frontends/qt4/ui/ThesaurusUi.ui:138 src/frontends/qt4/ui/WrapUi.ui:173 > < #: src/frontends/qt4/GuiParagraph.cpp:163 > src/frontends/qt4/GuiView.cpp:2737 --- >> #: src/frontends/qt4/ui/InsetParamsUi.ui:138 >> #: src/frontends/qt4/ui/LogUi.ui:88 >> #: src/frontends/qt4/ui/ParagraphUi.ui:261 >> #: src/frontends/qt4/ui/PrefsUi.ui:90 src/frontends/qt4/ui/RefUi.ui:343 >> #: src/frontends/qt4/ui/SearchUi.ui:203 >> #: src/frontends/qt4/ui/ShowFileUi.ui:55 >> #: src/frontends/qt4/ui/SymbolsUi.ui:66 >> #: src/frontends/qt4/ui/TexinfoUi.ui:58 >> #: src/frontends/qt4/ui/ThesaurusUi.ui:138 >> #: src/frontends/qt4/ui/WrapUi.ui:173 >> #: src/frontends/qt4/GuiParagraph.cpp:163 >> #: src/frontends/qt4/GuiView.cpp:2737 > ... > > Using msguniq with defined smaler page width does not help, because the > following call to xgettext changes the created .pot file again. > I did not find a parameter for xgettext to modify the line length of the > pot-file. xgettext does not reformat existing comment lines on my machine. It only adds comments for the files it did extract, and if the last existing comment line is short enough, it fills it up, for all remaining files to be added it creates new comment lines. Therefore, I believe that xgettext cannot be made responsible for differences in comment lines containing only .ui or .layout files. > Using the width param for msgmerge helps a little, but not all differences > are gone, instead new differences are created. I am out of ideas. I think that we did understand this particular difference now. fr.po in 2.3- staging was remerged by Uwe on windows, and there are formatting differences between gettext and msguniq, maybe because we feed files with windows line ends to gettext, maybe also for some other reason. Georg
