GNOME Bugzilla – Bug 749268
[Regression] 1.22 breaks user <note/>-s
Last modified: 2015-05-15 10:16:59 UTC
evolution-data-server contains custom gtk-doc code, like <note></note> [1], written exactly the same the guide at [2] suggests, but the new gtk-doc 1.22 fails to build HTML with it. Furthermore, it's gtk-doc 1.22 itself breaking the structure. The thing is that the generated .xml file has <para><note></para>, which breaks the XML parser. You can try to build evolution-data-server docs to get more similar oddities, which block it from building against 1.22. gtk-doc 1.21 doesn't have this issue. [1] https://git.gnome.org/browse/evolution-data-server/tree/libedataserver/e-source-webdav.c?h=gnome-3-16#n608 [2] https://developer.gnome.org/gtk-doc-manual/1.20/documenting_docbook.html.en
Isn't this a dupe of #749142 - can you try todays git?
(In reply to Stefan Sauer (gstreamer, gtkdoc dev) from comment #1) > Isn't this a dupe of #749142 - can you try todays git? (side note: if you prefix the bug number with a 'bug' word, then it'll be clickable and easier accessible, like here: bug #749142) gtk-doc git master at commit 0ea674a lets me build the gtk-doc for evolution-data-server, but still has some kind of regression of too many warnings like: > Warning: multiple "IDs" for constraint linkend: api-index-3.2. > Warning: multiple "IDs" for constraint linkend: api-index-3.12. > Warning: multiple "IDs" for constraint linkend: api-index-2.26. which are new here. the CPU usage is significantly higher too, noticed when it was printing these false warnings.
The warnings are new because gtk-doc now links to the api-index from since tags. For that it needs to give the index an id. It cannot rely on the user giving it an id like eds does https://git.gnome.org/browse/evolution-data-server/tree/docs/reference/eds/eds-docs.sgml#n297 You can either ignore the warning, or drop the id tags from the index in your master doc. For the CPU usage, do you have before/after figures? E.g. doing make clean; time make docs; a few times? There are many changes in the release, but nothing that would explain a significant slowdown.
I looked at some other projects and it seems to be common to use id="api-index-3-12" on the index tags, allthough its like 60:30. From my jhbuild checkout: > find . -name "*-docs.*ml" -exec grep "<index id=\"[a-z-]*[0-9-]*\"" {} \; | wc -l 344 > find . -name "*-docs.*ml" -exec grep "<index id=\"[a-z-]*[0-9.]*\"" {} \; | wc -l 254
Do you mean that eds should use dash in the version, instead of the dot? That's easily doable, if it's the way it should be done. The time difference is very significant, that I run it only once. With git master of gtk-doc I get: $ cd docs && make clean && time make real 1m34.108s user 1m32.724s sys 0m1.445s and with gtk-doc-1.21-1.fc21.noarch: $ cd docs && make clean && time make real 1m29.596s user 1m28.112s sys 0m1.513s The difference felt longer than 5 seconds, somehow. I changed the eds-docs.sgml to use api-index-X-YZ, but it still produces the warnings. This is out of scope of this bug report, thus I'd simply ignore this. Thanks and feel free to close this as a duplicate of bug #749142.
I run a git bisect. We loose 2.5 seconds due to commit 6d82b2651bac715be19a4d5fbb03052a6c000549. Hard to believe but I double checked. I filed bug #749419 to track it, but I think I'll release 1.23 without that being fixed, as it might be not trivial. Regarding the duplicated id's - changing the '.' for '-' in the ids worked for me, could you please check again? I can also submit that fix if you want. I'll mark this as duplicate for now. *** This bug has been marked as a duplicate of bug 749142 ***