Ross Burton : webkit-gtk: remove no-parallel bodge
A message that is not in the archive (sent off-list or not yet received): 20120718133146.0C2B910217@opal
Quoted 15 lines · On Wed, Jul 18, 2012 at 01:31:45PM +0000, [email protected] wrote:
Module: openembedded-core.git Branch: master Commit: 90ca88a6f9dd5e2bc055739f84cb224f9a82455a URL: http://git.openembedded.org/?p=openembedded-core.git&a=commit;h=90ca88a6f9dd5e2bc055739f84cb224f9a82455aAuthor: Ross Burton <[email protected]> Date: Tue Jul 17 19:44:19 2012 +0100webkit-gtk: remove no-parallel bodge
This workaround was originally for GNU Make 3.82 which upstream was blaming for a parallel build failure. Since then I've changed numerous variables and now parallel builds are working with 3.82 (and succeed in 20 minutes instead of 2 hours).
Quoted 5 lines
Revert the change until we've more data about what causes the breakage, then we can bring it back with more finesse.
Signed-off-by Ross Burton <[email protected]>Signed-off-by Saul Wold <[email protected]>
Now it failed on one of my builders:
| arm-oe-linux-gnueabi-libtool: link: arm-oe-linux-gnueabi-ranlib .libs/libWebCoreGtk.a
| mv -f Tools/DumpRenderTree/gtk/.deps/Programs_DumpRenderTree-TextInputController.Tpo Tools/DumpRenderTree/gtk/.deps/Programs_DumpRenderTree-TextInputController.Po | arm-oe-linux-gnueabi-libtool: link: ( cd "TestNetscapePlugin/.libs" && rm -f "libtestnetscapeplugin.la" && ln -s "../libtestnetscapeplugin.la" "libtestnetscapeplugin.la"
)
| arm-oe-linux-gnueabi-libtool: link: cannot find the library libjavascriptcoregtk-1.0.la' or unhandled argument libjavascriptcoregtk-1.0.la'
| make[1]: *** [Programs/minidom] Error 1
| arm-oe-linux-gnueabi-libtool: link: ( cd ".libs" && rm -f "libWebCoreGtk.la" && ln -s "../libWebCoreGtk.la" "libWebCoreGtk.la" )
GNU Make 3.82
I've built it without PARALLEL_MAKE about 8 times already and this is first time it failed.. so the race condition is probably still there.
Cheers,
Signature · 1 line
Martin 'JaMa' Jansa jabber: [email protected]
Quoted 2 lines · Martin Jansa
I've built it without
PARALLEL_MAKEabout 8 times already and this is first time it failed.. so the race condition is probably still there.
Disabling parallel makes really does make webkit take a long time to build. My preferred option at the moment is to make webkit-gtk depend on make-native-replacement, and package 3.81. I'd like to see that dependency only kick in it we detect that the host has 3.82, but that's pretty evil...
I've also pinged the bug-make mailing list about this problem as there have been a bug and two patches in their bug tracker for two years now with no comment.
Ross
Quoted 27 lines · Martin Jansa
Quoted 7 lines
This workaround was originally for GNU Make 3.82 which upstream was blaming for a parallel build failure. Since then I've changed numerous variables and now parallel builds are working with 3.82 (and succeed in 20 minutes instead of 2 hours).
Revert the change until we've more data about what causes the breakage, then we can bring it back with more finesse.
Now it failed on one of my builders:
| arm-oe-linux-gnueabi-libtool: link: arm-oe-linux-gnueabi-ranlib .libs/libWebCoreGtk.a
| mv -f Tools/DumpRenderTree/gtk/.deps/Programs_DumpRenderTree-TextInputController.Tpo Tools/DumpRenderTree/gtk/.deps/Programs_DumpRenderTree-TextInputController.Po | arm-oe-linux-gnueabi-libtool: link: ( cd "TestNetscapePlugin/.libs" && rm -f "libtestnetscapeplugin.la" && ln -s "../libtestnetscapeplugin.la" "libtestnetscapeplugin.la"
)
| arm-oe-linux-gnueabi-libtool: link: cannot find the librarylibjavascriptcoregtk-1.0.la' or unhandled argumentlibjavascriptcoregtk-1.0.la'
| make[1]: *** [Programs/minidom] Error 1
| arm-oe-linux-gnueabi-libtool: link: ( cd ".libs" && rm -f "libWebCoreGtk.la" && ln -s "../libWebCoreGtk.la" "libWebCoreGtk.la" )GNU Make 3.82
Quoted 7 lines
I've built it without
PARALLEL_MAKEabout 8 times already and this is first time it failed.. so the race condition is probably still there.Cheers,
--
Martin 'JaMa' Jansa jabber: [email protected]
I have tried 3 times (with cleansstate) and got 100% failure - similar to Martin's:
| arm-angstrom-linux-gnueabi-libtool: link: cannot find the library libjavascriptcoregtk-1.0.la' or unhandled argument libjavascriptcoregtk-1.0.la'
| make[1]: *** [Programs/unittests/testdomdomwindow] Error 1 | make[1]: *** Waiting for unfinished jobs.... | arm-angstrom-linux-gnueabi-libtool: link: cannot find the library
libjavascriptcoregtk-1.0.la' or unhandled argument libjavascriptcoregtk-1.0.la'
| make[1]: *** [Programs/unittests/testdomdocument] Error 1
I reverted this patch and build is fine.
Andreas
Quoted 11 lines · Andreas Müller
| arm-angstrom-linux-gnueabi-libtool: link: cannot find the library
libjavascriptcoregtk-1.0.la' or unhandled argumentlibjavascriptcoregtk-1.0.la'| make[1]: *** [Programs/unittests/testdomdomwindow] Error 1 | make[1]: *** Waiting for unfinished jobs.... | arm-angstrom-linux-gnueabi-libtool: link: cannot find the library
libjavascriptcoregtk-1.0.la' or unhandled argumentlibjavascriptcoregtk-1.0.la'
| make[1]: *** [Programs/unittests/testdomdocument] Error 1I reverted this patch and build is fine.
Bringing this thread back as it needs to get resolved.
Can you try adding the patch in
https://bug-79498-attachments.webkit.org/attachment.cgi?id=131681 to webkit and seeing if it builds with -j then? That patch claims to add duplicated rules to force the buggy make not to drop essential dependencies.
If that doesn't help... I don't know what the best thing to do is. Add a make-replacement 3.81 recipe, and depend on it? Detect the make version and disable parallel and warn if it's 3.82?
Cheers,
Ross