GNOME Bugzilla – Bug 584802
gtk-doc scanner crashes
Last modified: 2012-03-15 09:16:34 UTC
gnome-keyring-2.26.1 on OS X 10.4 with gtkdoc-1.11. I use --enable-gtk-doc, and 'make' eventually fails in docs/reference/gcr: gtk-doc: Running scanner gcr-scan gck-rpc-module.c:228: failed assertion `pkcs11_socket_path[0]' Scan failed: make[4]: *** [scan-build.stamp] Error 255 However, reference/gnome-keyring and reference/gp11 gtk-doc all builds successfully. I hacked some debugging into gcr-scan.c, and traced the problem to when get_object_types() calls g_type_class_ref(object_types[i]) for object_types[i]=gcr_parser_get_type() g-k-2.24.1 was not broken (but I don't know how it was built, just that it is present on my system).
Same bug on gentoo: http://bugs.gentoo.org/show_bug.cgi?id=267957
Very strange, what is gtk-scan doing running code inside of the rpc-module. That module doesn't even use GObject. In any case I need a full backtrace from the assertion. Could you (or the gentoo guys who can duplicate this) run the gcr-scan command (with full arguments) in a debugger and get a backtrace? Thanks!
Presumably cured via Bug #553164. However, the double-build issue means there is still a problem: if an older gnome-keyring (with the stray assert) is installed at same prefix as for where one is building a new one, the module-loader apparently loads the installed module so the old-but-now-fixed-bug still bugs. That's broken behavior to see installed old copy while building new. Somehow gcr-scan needs to be able to disable module-loading or something like that.
gcr-scan is gtkdoc code. I have no idea where to go from here...
Haven't seen/heard about this in a couple years no, so assuming its fixed.