Bug 76161

I cannot reproduce this one. Do you mind deleting WebKitBuild/Dependencies and trying again?

(In reply to comment #1) > I cannot reproduce this one. Do you mind deleting WebKitBuild/Dependencies and trying again? Did it clones again jhbuild and keep looping for ever until the death of my machine.

It clones in WebKitBuild in my source dir, maybe it shouldn't "pollute" the source dir.

(In reply to comment #3) > It clones in WebKitBuild in my source dir, maybe it shouldn't "pollute" the source dir. That's true, but this is unrelated to the bug. Do you mind trying to install the gnome-common package (or whatever brings in /usr/bin/gnome-autogen.sh) and trying again?

Same but I found suspicious this line : Don't forget to create ~/.jhbuildrc install -m755 install-check /home/darktears/dev/troll//qt-50-debug/qtbase/bin/install-check why it would put stuff in my Qt installation??? Just for the record it is my first entry in PATH.

(In reply to comment #5) > why it would put stuff in my Qt installation??? Just for the record it is my first entry in PATH. Perhaps this is a bug in the fallback jhbuild build script. It shouldn't be installing anything outside the prefix.

*** Bug 80698 has been marked as a duplicate of this bug. ***

(In reply to comment #7) > *** Bug 80698 has been marked as a duplicate of this bug. *** /home/darktears/dev/troll/webkit/WebKitBuild/Dependencies/Source/jhbuild/jhbuild is where it is cloned.

(In reply to comment #8) > (In reply to comment #7) > > *** Bug 80698 has been marked as a duplicate of this bug. *** > > /home/darktears/dev/troll/webkit/WebKitBuild/Dependencies/Source/jhbuild/jhbuild is where it is cloned. Also gnome-common seems to be a hard dependency, the fallback script of jhbuild without autotool is proper broken.

Created attachment 138178 [details] Patch

Should we add gnome-common to jhbuild, or do you think it would be better as a dependency we expect the system to provide?

I added gnome-common as a dependency on the wiki.

Comment on attachment 138178 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=138178&action=review Great! There are just a few minor change suggestions below. > Tools/jhbuild/jhbuild-wrapper:48 > +if os.environ.has_key('WEBKITOUTPUTDIR'): > + installation_prefix = os.path.abspath(os.path.join(os.environ['WEBKITOUTPUTDIR'], 'Dependencies', 'Root')) > + source_path = os.path.abspath(os.path.join(os.environ['WEBKITOUTPUTDIR'], 'Dependencies', 'Source')) > + jhbuild_source_path = os.path.join(source_path, 'jhbuild') > + jhbuild_path = os.path.join(installation_prefix, 'bin', 'jhbuild') > +else: > + installation_prefix = os.path.abspath(top_level_path('WebKitBuild', 'Dependencies', 'Root')) > + source_path = os.path.abspath(top_level_path('WebKitBuild', 'Dependencies', 'Source')) > + jhbuild_source_path = os.path.join(source_path, 'jhbuild') > + jhbuild_path = top_level_path('WebKitBuild', 'Dependencies', 'Root', 'bin', 'jhbuild') > + I think in this case something like this might be better: if os.environ.has_key('WEBKITOUTPUTDIR'): dependencies_path = os.path.abspath(top_level_path('WebKitBuild', 'Dependencies')) else: dependencies_path = top_level_path('WebKitBuild', 'Dependencies') And then you can make all paths relative to those. I think a change like this makes sense for the previous two code blocks too.

*** Bug 81475 has been marked as a duplicate of this bug. ***

Committed r115532: <http://trac.webkit.org/changeset/115532>

(In reply to comment #15) > Committed r115532: <http://trac.webkit.org/changeset/115532> So this patch restored the behavior to store the jhbuild*.md5sum files in the build directory instead of Dependencies. Is there any reason why?

(In reply to comment #16) > (In reply to comment #15) > > Committed r115532: <http://trac.webkit.org/changeset/115532> > > So this patch restored the behavior to store the jhbuild*.md5sum files in the build directory instead of Dependencies. Is there any reason why? Hum doesn't seem to be this one. Anyway, something broke :( Will try to track this regression down again.

Comment on attachment 138178 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=138178&action=review > Tools/Scripts/webkitdirs.pm:1846 > + return join(baseProductDir(), "Dependencies"); Hrm. the first argument to join() should be the separator no?? > Tools/Scripts/webkitdirs.pm:1991 > + my $destination = join(getJhbuildPath(), $file); Ditto.

Comment on attachment 138178 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=138178&action=review >> Tools/jhbuild/jhbuild-wrapper:48 >> + > > I think in this case something like this might be better: > > if os.environ.has_key('WEBKITOUTPUTDIR'): > dependencies_path = os.path.abspath(top_level_path('WebKitBuild', 'Dependencies')) > else: > dependencies_path = top_level_path('WebKitBuild', 'Dependencies') > > And then you can make all paths relative to those. I think a change like this makes sense for the previous two code blocks too. Looks like this suggestion wasn't correctly applied :( Current code: if os.environ.has_key('WEBKITOUTPUTDIR'): dependencies_path = os.path.abspath(os.path.join(os.environ['WEBKITOUTPUTDIR'], 'Dependencies', 'Root')) else: dependencies_path = os.path.abspath(top_level_path('WebKitBuild', 'Dependencies', 'Root')) so we end up with installation_prefix ending with Root/Root ... Will send another patch.

Oh this was fixored in r115904... My bad for not updating before blaming :)

Wow, and I thought I had tested it properly =) thanks for your investigations, hopefully it's all fixed now.