After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 640241 - non-srcdir builds busted
non-srcdir builds busted
Status: RESOLVED FIXED
Product: gtk-doc
Classification: Platform
Component: general
unspecified
Other Linux
: Normal normal
: 1.17
Assigned To: gtk-doc maintainers
gtk-doc maintainers
Depends on:
Blocks: 127049
 
 
Reported: 2011-01-22 05:42 UTC by Matthias Clasen
Modified: 2011-02-22 14:02 UTC
See Also:
GNOME target: ---
GNOME version: ---



Description Matthias Clasen 2011-01-22 05:42:44 UTC
With 1.16, my distchecks fail. Looking closer, it turns out that .e.g the gio overview.xml file does not get copied over. I think the setup-build.stamp rule is faulty.

‣‧‧‧‧‧‧‧           test -f $(abs_srcdir)/$$file || \
‣‧‧‧‧‧‧‧               cp -a $(abs_srcdir)/$$file $(abs_builddir)/; \

should be 

‣‧‧‧‧‧‧‧           test -f $(abs_builddir)/$$file || \
‣‧‧‧‧‧‧‧               cp -a $(abs_srcdir)/$$file $(abs_builddir)/; \
Comment 1 Matthias Clasen 2011-01-22 13:17:46 UTC
It seems that non-srcdir builds are more thoroughly busted than that.

Just try to distcheck glib.
Comment 2 Stefan Sauer (gstreamer, gtkdoc dev) 2011-01-23 11:27:22 UTC
First, the out-of-srcdir build was never meant to be part of the relase. Dunno how it end up being committed. Too bad that no one tests pre-releases also. I have been successfully testing glib with it. I will do that again at some point.
Comment 3 Stefan Sauer (gstreamer, gtkdoc dev) 2011-01-23 13:33:58 UTC
I can't fully test "make distcheck" for srcdir=builddir right now - the doc scanner for gio segfaults.
export GTK_DOC_KEEP_INTERMEDIATE=1
cd glib-2.27.93/_build/docs/reference/gio
gdb ./gio-scan
...
Program received signal SIGSEGV, Segmentation fault.
g_io_extension_get_type (extension=0x50) at ../../gio/giomodule.c:905
905	{
(gdb) bt
  • #0 g_io_extension_get_type
    at ../../gio/giomodule.c line 905
  • #1 get_object_types
    at gio-scan.c line 186
  • #2 main
    at gio-scan.c line 308

Comment 4 Stefan Sauer (gstreamer, gtkdoc dev) 2011-01-27 14:10:03 UTC
Just confirming on a different machine the glib git build is indeed broken:

libtool: link: gcc -g -O2 -Wall .libs/gio-scan.o -o .libs/gio-scan  ../../../glib/.libs/libglib-2.0.so ../../../gobject/.libs/libgobject-2.0.so ../../../gmodule/.libs/libgmodule-2.0.so ../../../gio/.libs/libgio-2.0.so -Wl,-rpath -Wl,/home/ensonic/debug/lib
.libs/gio-scan.o: In function `get_object_types':
/home/ensonic/projects/gnome/glib/_debug/docs/reference/gio/gio-scan.c:193: undefined reference to `g_registry_backend_get_type'
collect2: ld returned 1 exit status
Linking of scanner failed:
Comment 5 Stefan Sauer (gstreamer, gtkdoc dev) 2011-01-27 14:12:02 UTC
Before someone brings it up - build.gnome.org is not affected:
...
checking whether to build gtk-doc documentation... no
...
Comment 6 Stefan Sauer (gstreamer, gtkdoc dev) 2011-01-27 15:15:36 UTC
Sorry posted the wrong one: It also dies at the same point:
Program received signal SIGSEGV, Segmentation fault.
g_io_extension_get_type (extension=0x1) at ../../gio/giomodule.c:904
904	g_io_extension_get_type (GIOExtension *extension)
(gdb) bt
  • #0 g_io_extension_get_type
    at ../../gio/giomodule.c line 904
  • #1 __libc_start_main
    at libc-start.c line 226
  • #2 _start

and this was caused by an leftover in .types. But thats fixed in glib git already.

I pushed one commit to gtk-doc git to fix that issue.
Comment 7 Stefan Sauer (gstreamer, gtkdoc dev) 2011-01-27 15:26:30 UTC
Should be working in gtk-doc git head. It was working for non-tmpl based projects before.
Comment 8 Stefan Sauer (gstreamer, gtkdoc dev) 2011-02-01 23:19:18 UTC
Just to give status after several evenings of working:
1.) I still face leftover files when building the docs for glib.

2.) grep "Computing" make.log | wc -l
18
That is each doc module is build 6 times instead of 2 times! Once its happening because of "make check", but I don't know why yet. If I can't figure it in the coming 3 days I back-out the patch and release a 1.17 with the previous Makefiles.
Comment 9 Stefan Sauer (gstreamer, gtkdoc dev) 2011-02-01 23:40:31 UTC
quick update:
2.) grep "Computing" make.log | wc -l
9
Thats still 1 round too much, but I have hope again.
Comment 10 Vincent Untz 2011-02-02 00:43:39 UTC
FWIW, you can try distcheck on libwnck master if you want something faster than glib ;-) Not sure it's the same issue, but it fails too.
Comment 11 Stefan Sauer (gstreamer, gtkdoc dev) 2011-02-02 08:29:47 UTC
(In reply to comment #10)
> FWIW, you can try distcheck on libwnck master if you want something faster than
> glib ;-) Not sure it's the same issue, but it fails too.

Having something that is small, affected, but does not require gtk+-3 would be nice.
Comment 12 Vincent Untz 2011-02-02 10:34:35 UTC
libwnck from gnome-2-30 also has the issue, if you don't want gtk+-3.
Comment 13 Stefan Sauer (gstreamer, gtkdoc dev) 2011-02-02 22:19:19 UTC
==============================================
glib-2.27.94 archives ready for distribution: 
glib-2.27.94.tar.gz
==============================================

So please, people try again.
Comment 14 Stefan Sauer (gstreamer, gtkdoc dev) 2011-02-22 14:02:48 UTC
For #gtk+ irc.

14:50 < ensonic> mclasen: is make distcheck working fine for you now?
14:50 < mclasen> ensonic: it was working the last time I tried, thanks