"
USE_WXNEEDED, add "-Wl,-z,wxneeded" to pkg-config files
WebKit violates W^X, so binaries linked with it need to be "wxneeded".
Pkg-config files seem to be an appropriate mean of automating this task.
"
that time to workaround a gmake weirdness triggering forkbombs, but now
apparently it causes weird build failures with binutils error messages
when using MAKE_JOBS>=4, as experienced by sthen@, naddy@ & espie@.
Unfortunately, i don't have hardware with 4 cores or more, so didnt see
those... Removing the chunk doesnt seem to show gmake weirdness anymore,
so just do it (tm). Analysed by espie@.
Note that gtk3 FLAVOR is broken on powerpc with "undefined reference to
`__atomic_fetch_add_8'". /me goes crying in a corner.
building webkit is hard, let's go shopping.
See http://webkitgtk.org/2014/03/24/webkitgtk2.4.0-released.html. Note
that Webkit2 API is still broken, on OpenBSD(and FreeBSD) and this might
be the last release shipping the Webkit1 API.
Pass -DLOG_DISABLED=0 to CPPFLAGS to somewhat allow debugging.
Use the same ruby version on sparc64 as other archs now that they're
more reliable.
Remove some patches merged upstream, and add links to upstream bz for
most of the remaining patches.
Note: this needs gnu/usr.bin/binutils/include/opcode/i386.h r1.15 to
build. And gcc 4.8 (and python, ruby & perl :)
Built numerous times on amd64/ppc/sparc64 (broken at runtime on the last
two, but that's not a regression from 2.2..) and went in a bunch of
bulks.
- Now depends on newer gcc, use 4.6 by default.
- Enable webkit2 api in the gtk3 FLAVOR. Still broken/erratic behaviour at
runtime, but at least this will allow people to debug it, and it's
required by newer GNOMEs anyway. Provide MiniBrowser for debugging purposes.
- add hack from bug #118732 to workaround the fact that our binutils
doesnt support 'ar T'
- update patches from #103128 to tentatively fix powerpc runtime
- force-disable YARR_JIT & ASSEMBLE on archs without JIT (ie anything
besides i386/amd64) via Platform.h patch instead of JSC_CPPFLAGS (which
doesnt work anymore anyway)
- add a bunch of patches to fix build (missing includes etc) in the WebKit2
source tree
Tested in an amd64 bulk build, runtime-tested previous betas on
i386/amd64, test-built on powerpc and sparc64. Things might still be
broken at runtime on powerpc, but at least it allows more GNOME things
to be updated.
pushed upstream as https://bugs.webkit.org/show_bug.cgi?id=114978)
Seems to fix xombrero/surf crashes on amd64, and seems to magically make
webkit sort-of usable (ie dead-slow, but js seems to work) on powerpc.
(symptom : ** Message: console message: undefined @0: RangeError: Maximum
call stack size exceeded.)
Note : webkit still fails to build on mips64* and hppa. sigh.
ok jasper@
- add ruby to BDEPS. Webkit build system now uses autohell, python, ruby and perl \o/
- set MODRUBY_REV = 1.8 on sparc64 since 1.9 is broken there
- disable WebKit2 support, it needs shm_open()/shm_unlink()
- switch from gstreamer-0.10 to gstreamer1
- force-disable JIT/ASSEMBLER on non-jit archs, builds fails otherwise (bug #113638)
- only build NumberPrototype.cpp with -O0 on sparc64, not the damn whole thing
- use -Wl,--relax on ppc otherwise linking fails
- move some patches around since JavaScriptCore/wtf moved to WTF/wtf
- dont uselessly rerun make within make during install, gmake blows
- blindly backport upstream r144077 to fix mips64 build
- backport bits of #103128 & r145551 to fix powerpc build
- backport r147967 & more to fix bashisms in configure
- bug #109932 forced the use of gcc >= 4.7 or clang >= 3.0, but this is
only for WebKit2 so loosen the check atm
Tested working fine on amd64 & i386, went in bulk builds, known broken
at runtime on ppc (being worked on. or not.) should build on mips64el
(and maybe even run!) and sparc64 (wont run! ever! nothing new here..),
but 'someone' will care about the poor exotic archs and fix'em, right ?
After too much prodding from the gnome maintainers who desperately need it
- disable static libs, shrinks package size by a factor 3
- now depends on geoclue for geolocation purposes (of course its up to
the browsers to enable the feature or not..)
- simplify Makefile/FLAVORS/SHARED_LIBS
- fold most of the PFRAGs to PLIST by using variables
- remove patch-Source_JavaScriptCore_runtime_JSValue_h, seems it was
added for sparc64 but now it breaks the build there, and noone has been
able to justify the addition of that patch. Without it, GtkLauncher
shows simple html websites on sparc64 and crashes as soon as some js is
involved, but thats not much different from before
- remove patch-Source_JavaScriptCore_wtf_text_StringHash_h &
patch-Source_JavaScriptCore_wtf_text_StringImpl_cpp, replaced by
patch-Source_JavaScriptCore_runtime_UString_h for mips64/sparc64
- add patch-Source_JavaScriptCore_wtf_NumberOfCores_cpp for #82585
tested on i386/amd64 by several, amd64 bulk build, ppc & sparc64 testing
by myself. mips64el in the works..
ok ajacoutot@
