Created attachment 182554 [details] v0 I've been dogfooding 17.1.0 on i965 (drm-next, Skylake) since rc1 without issues. poudriere bulk -t is green on 10.3 i386/amd64, 11.0 i386/amd64, 12.0C amd64. ABI: https://abi-laboratory.pro/tracker/timeline/mesa/ (exp-run is unnecessary)
*** Bug 218551 has been marked as a duplicate of this bug. ***
Comment on attachment 182554 [details] v0 > + --enable-llvm-shared-libs Oops, rebase cruft.
Comment on attachment 182554 [details] v0 files/configure.ac + files/patch-configure bits are to be reviewed via https://lists.freedesktop.org/archives/mesa-dev/2017-May/155485.html
Created attachment 182561 [details] v1 Better fix for missing __atomic_fetch_add_8 on 11.0 i386 based on upstream feedback. I have no clue why 10.3 i386 isn't affected.
Created attachment 182651 [details] v2 - Avoid FREEDRENO dependening on textproc/py-mako - Fix VC4 similar to libsync.h
Created attachment 182652 [details] v2.1 Fix plist on aarch64 Error: Orphaned: lib/dri/kgsl_dri.so Error: Orphaned: lib/dri/msm_dri.so Error: Orphaned: lib/dri/vc4_dri.so
A commit references this bug: Author: rezny Date: Sun May 21 19:30:56 UTC 2017 New revision: 441424 URL: https://svnweb.freebsd.org/changeset/ports/441424 Log: Update the Mesa ports to 17.1.0 * restrict use of LLVM for Gallium to platforms with LLVM in base to fix runtime errors when Gallium is built with a mix of GCC and LLVM * make LLVM only a BUILD_DEPEND for mesa-libs, which needs it to build EGL * add pkg-plist entries for freedreno and vc4 drivers [1] * remove extraneous SHEBANG_FILES entry to fix build of freedreno [2,3] * correct typo in libosmesa, which may have affected the Gallium build [4] PR: 219247 [1], 219406 [3] Submitted by: jbeich [1] (2 patches, pkg-plist fix), strejda (vc4 CFLAGS) Reported by: strejda [2], andrew [3], linimon [4] Approved by: swills (mentor) Differential Revision: https://reviews.freebsd.org/D10842 Changes: head/graphics/libosmesa/Makefile head/graphics/mesa-dri/Makefile head/graphics/mesa-dri/Makefile.common head/graphics/mesa-dri/Makefile.targets head/graphics/mesa-dri/distinfo head/graphics/mesa-dri/files/configure.ac head/graphics/mesa-dri/files/patch-configure head/graphics/mesa-dri/files/patch-src_egl_drivers_dri2_platform__x11.c head/graphics/mesa-dri/files/patch-src_gallium_auxiliary_pipe-loader_pipe__loader__drm.c head/graphics/mesa-dri/files/patch-src_gallium_drivers_vc4_Makefile.in head/graphics/mesa-dri/files/patch-src_gallium_drivers_vc4_vc4__bufmgr.c head/graphics/mesa-dri/files/patch-src_intel_tools_aubinator.c head/graphics/mesa-dri/files/patch-src_intel_tools_aubinator__error__decode.c head/graphics/mesa-dri/files/patch-src_mapi_glapi_gen_gl__gentable.py head/graphics/mesa-dri/files/patch-src_mesa_main_compiler.h head/graphics/mesa-dri/files/patch-src_util_build__id.c head/graphics/mesa-dri/files/patch-src_util_ralloc.c head/graphics/mesa-dri/pkg-plist head/graphics/mesa-libs/Makefile head/lang/clover/Makefile
I was unaware you were tracking Mesa development until I found this PR. Perhaps we should be more closely collaborating. If you're on IRC, drop by #freebsd-xorg sometime. I used a couple of your simple patches and of course the plist corrections in the 17.1 update, but I already had simpler fixes for the other build issues. The one-line fix for GCC atomics should give a clue why it was only an issue on 11.0 i386 and not 10.3 (where I'd already worked around that issue).
(In reply to Matthew Rezny from comment #8) > I was unaware you were tracking Mesa development until I found this PR. Perhaps > we should be more closely collaborating. If you're on IRC, drop by > #freebsd-xorg sometime. I do collaborate but mainly via mailing lists and bugzilla. IRC has too many downsides (no Tor support, lack of public logs, requires stable connection, different timezones) to make it hard to keep up. > I used a couple of your simple patches and of course the plist corrections in > the 17.1 update, but I already had simpler fixes for the other build issues. > The one-line fix for GCC atomics should give a clue why it was only an issue on > 11.0 i386 and not 10.3 (where I'd already worked around that issue). I don't like your fix for USES=shebangfix. Many files are still pointlessly regenerated. Maybe drop the whole SHEBANG_FILES as **/Makefile.am prepends $(PYTHON_GEN) which is more portable or (like in my patch) reset timestamps.