The aec3 module attempts to do SSE2 runtime detection in a way that does
not compile with GCC, so disable it. (SSE2 use in that module is now
compile-time only, i.e., only enabled for the x86_64 package in Fedora.)
The patch also optimizes x86_64 by making the SSE2 detection return a
compile-time constant, which upstream neglected doing.
- no-sse2 patch: First try at cleaning up the V8 x87 backend bitrot, by
porting needed changes from the ia32 SSE2 backend.
- Reduce debugging info on ARM also on F27+ (as on F26- since 5.9.0).
- Update changelog date.
* Revert https://chromium-review.googlesource.com/c/chromium/src/+/570351
which assumes that x86 must have SSE (1).
* Fix the wsola_internals.cc patch: xmmintrin.h requires only SSE (1),
not SSE2. (Should not really make a difference in practice though.)
* Fix rebasing copy&paste error in DirectConvolver.cpp.
Add missing libavutil/imgutils.c, libavutil/imgutils.h,
libavutil/imgutils_internal.h and libavcodec/hwaccel.h.
Also update sources with the regenerated and reuploaded tarball.
* Mon Dec 25 2017 Kevin Kofler <Kevin@tigcc.ticalc.org> - 5.10.0-1
- Update to 5.10.0
- Update version numbers of bundled stuff
- Drop support for Fedora < 26 (in particular, WEBENGINE_CONFIG F25 workarounds)
- Drop qt57 patch, support for Qt 5.7 was completely dropped upstream
- Update get_free_ffmpeg_source_files.py from Fedora Chromium packaging
- clean_qtwebengine.sh: Update for the changed tarball naming scheme
- Use QMAKE_EXTRA_ARGS instead of the removed WEBENGINE_CONFIG
- Rebase linux-pri, system-nspr-prtime, system-icu-utf, no-sse2, skia-neon and
gn-bootstrap-verbose patches
- In particular, restore the removed V8 x87 backend in the no-sse2 patch
- Re-backport no-aspirational-scripts from upstream (undo 5.9 backport)
- arm-fpu-fix patch: Drop the host tools hunk added in 5.9.0-2, fixed upstream
Also rename the patch file from
qtwebengine-opensource-src-5.9.0-arm-fpu-fix.patch to
qtwebengine-opensource-src-5.9.2-arm-fpu-fix.patch because the patch was
rebased. Otherwise, it is easy to miss that the patch was rebased. (I
did, I only noticed it now while working on the RPM Fusion package. And
I might easily have missed it again and run into a build failure due to
that.)
* Sat Oct 14 2017 Kevin Kofler <Kevin@tigcc.ticalc.org> - 5.9.2-2
- Do not call the Chromium unbundling script on re2, QtWebEngine now
auto-detects and uses the system re2 out of the box
- Drop system-re2 patch (patching the no longer used unbundle/re2.gn), the
QtWebEngine re2/BUILD.gn is already correct
* Fri Jun 30 2017 Kevin Kofler <Kevin@tigcc.ticalc.org> - 5.9.1-1
- Update to 5.9.1
- Rebase qtbug-61521 patch (drop the parts that are already in 5.9.1)
- Drop backported GN aarch64 patches already included in 5.9.1
* Sun Jun 25 2017 Kevin Kofler <Kevin@tigcc.ticalc.org> - 5.9.0-3
- Fix broken binary compatibility for C memory management functions (incomplete
upstream fix for QTBUG-60565)
* Tue Jun 13 2017 Kevin Kofler <Kevin@tigcc.ticalc.org> - 5.9.0-2
- arm-fpu-fix patch: Also build the host tools (i.e., GN) with the correct FPU
This ensures GN does not get built with NEON on 32-bit armv7hl. The ARM
builders don't necessarily support NEON. (The Fedora one apparently
does, but the RPM Fusion one apparently does not.)
