Bug 270348
| Summary: | graphics/openexr: build fails on armv7 | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | Ports & Packages | Reporter: | Robert Clausecker <fuz> | ||||||||
| Component: | Individual Port(s) | Assignee: | Matthias Andree <mandree> | ||||||||
| Status: | Closed FIXED | ||||||||||
| Severity: | Affects Some People | Keywords: | regression | ||||||||
| Priority: | --- | Flags: | mandree:
maintainer-feedback+
|
||||||||
| Version: | Latest | ||||||||||
| Hardware: | arm | ||||||||||
| OS: | Any | ||||||||||
| See Also: |
https://github.com/AcademySoftwareFoundation/openexr/issues/1365 https://github.com/AcademySoftwareFoundation/openexr/issues/1367 https://github.com/AcademySoftwareFoundation/openexr/pull/1368 |
||||||||||
| Attachments: |
|
||||||||||
Robert, thanks for the report. I'll poke at it. Looks like a genuine upstream bug introduced into 3.1.6 with https://github.com/AcademySoftwareFoundation/openexr/pull/1348 or https://github.com/AcademySoftwareFoundation/openexr/commit/677c6a52d8cde3b8630932a93b631d2d4e68ab52 This seems to only branch on __ARM_NEON without looking further if it's on __aarch64 or not. I have forwarded this upstream to https://github.com/AcademySoftwareFoundation/openexr/issues/1365 Robert: if I were to back out that particular commit only for ARMv7 in case they don't manage a fix in time, how would I word the if I am on "ARM but not AARCH64" in the port's Makefile? Created attachment 240995 [details]
try to fix ARMv7 compilation, patch #1 of 2
Created attachment 240996 [details]
try to fix ARMv7 compilation, patch #2 of 2
Robert, please put both patches into ports/graphics/openexr/files/ and recompile on ARMv7 to see if that fixes the issue. If you can, please also try on AArch64 and check if it still enables the reconstruct_neon and interleave_neon there.
A commit in branch main references this bug: URL: https://cgit.FreeBSD.org/ports/commit/?id=c185d09d3e1e2e3881284b8eca93028db5575c47 commit c185d09d3e1e2e3881284b8eca93028db5575c47 Author: Matthias Andree <mandree@FreeBSD.org> AuthorDate: 2023-03-19 23:26:53 +0000 Commit: Matthias Andree <mandree@FreeBSD.org> CommitDate: 2023-03-19 23:28:16 +0000 graphics/openexr: mark BROKEN on armv7 Upstream bug report: https://github.com/AcademySoftwareFoundation/openexr/issues/1365 PR: 270348 graphics/openexr/Makefile | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) Upstream project's contributor wrote he will follow a similar approach as proposed here later today. I'll go ahead and test your patch as soon as my armv7 box is done build-testing my current set of ports patches. Created attachment 241037 [details]
openexr-3.1.6 armv7 build log w/ attachments 24099{5,6}
The patch set you posted does not fix the build.
Please see attached build log.
Comment on attachment 241037 [details]
openexr-3.1.6 armv7 build log w/ attachments 24099{5,6}
Uh. Can you re-run with "make -k" so we trap all errors? ImfDwaCompressorSimd.h wasn't on my desk so far, and also isn't addressed by upstream.
I don't have time for additional runs today. Maybe tomorrow. I can give you access to an armv7 box for testing if you like (send me an email). I should be more precise, and the failures you are now posting seem to be related to https://github.com/AcademySoftwareFoundation/openexr/commit/436fcd2829ae9a8965af1db15ac8531fdc8b96ce I've sent ssh keys by direct mail to your FreeBSD address. DwaCompressor NEON on ARMv7 bugs filed as https://github.com/AcademySoftwareFoundation/openexr/issues/1367. I have local patches that, together with PR1366 from upstream, fix the ARMv7 builds, but I need to have a look at the ARMv8/AArch64 output files to make sure we don't void the AARCH64 (NEON) optimizations there. A commit in branch main references this bug: URL: https://cgit.FreeBSD.org/ports/commit/?id=54d6860aac464b690f37007cfaef52c5a0b427b0 commit 54d6860aac464b690f37007cfaef52c5a0b427b0 Author: Matthias Andree <mandree@FreeBSD.org> AuthorDate: 2023-03-21 16:42:19 +0000 Commit: Matthias Andree <mandree@FreeBSD.org> CommitDate: 2023-03-21 18:24:38 +0000 graphics/openexr: Fix ARMv7 build OpenEXR 3.1.6 introduced several NEON-based optimizations that implied Aarch64. Add patched, either picked from upstream, or written by mandree@, to enable those NEON features that also require Aarch64 only there. PR-1366 is cherry-picked from upstream, and patch-lib/patch-test files are my work but build upon said PR. Also cherry-pick PR1354 that adds a missing check for AVX, which is why I am bumping PORTREVISION because it might change code (I have not checked). https://github.com/AcademySoftwareFoundation/openexr/issues/1365 PR: 270348 Reported by: fuz@ (Robert Clausecker) graphics/openexr/Makefile | 6 +- ...-3f97750d1ec203e7d7eb8d5f30f3d5e7e68ad720 (new) | 25 ++++++ ...-a41a736d64e3d93baffef1042d4a3d1aaf74f1c9 (new) | 98 ++++++++++++++++++++++ .../patch-lib_OpenEXR_ImfDwaCompressor.cpp (new) | 11 +++ .../patch-lib_OpenEXR_ImfDwaCompressorSimd.h (new) | 29 +++++++ ...est_OpenEXRTest_testDwaCompressorSimd.cpp (new) | 38 +++++++++ 6 files changed, 203 insertions(+), 4 deletions(-) Local FreeBSD changes forwarded upstream as https://github.com/AcademySoftwareFoundation/openexr/pull/1368 or https://github.com/AcademySoftwareFoundation/openexr/commit/8d374c9e73a4867bacfbae401c34357243641ea0. files/patch-PR1366-* has obsoleted the two previous patches I had proposed, which had been a partial fix to the issue (as is upstream's pull request PR1366). A commit in branch 2023Q1 references this bug: URL: https://cgit.FreeBSD.org/ports/commit/?id=b7e05a72e644249c32c2e723d015b7af444c8690 commit b7e05a72e644249c32c2e723d015b7af444c8690 Author: Matthias Andree <mandree@FreeBSD.org> AuthorDate: 2023-03-21 16:42:19 +0000 Commit: Matthias Andree <mandree@FreeBSD.org> CommitDate: 2023-03-21 19:22:30 +0000 graphics/openexr: Fix ARMv7 build OpenEXR 3.1.6 introduced several NEON-based optimizations that implied Aarch64. Add patched, either picked from upstream, or written by mandree@, to enable those NEON features that also require Aarch64 only there. PR-1366 is cherry-picked from upstream, and patch-lib/patch-test files are my work but build upon said PR. Also cherry-pick PR1354 that adds a missing check for AVX, which is why I am bumping PORTREVISION because it might change code (I have not checked). https://github.com/AcademySoftwareFoundation/openexr/issues/1365 PR: 270348 Reported by: fuz@ (Robert Clausecker) (cherry picked from commit 54d6860aac464b690f37007cfaef52c5a0b427b0) graphics/openexr/Makefile | 4 +- ...-3f97750d1ec203e7d7eb8d5f30f3d5e7e68ad720 (new) | 25 ++++++ ...-a41a736d64e3d93baffef1042d4a3d1aaf74f1c9 (new) | 98 ++++++++++++++++++++++ .../patch-lib_OpenEXR_ImfDwaCompressor.cpp (new) | 11 +++ .../patch-lib_OpenEXR_ImfDwaCompressorSimd.h (new) | 29 +++++++ ...est_OpenEXRTest_testDwaCompressorSimd.cpp (new) | 38 +++++++++ 6 files changed, 203 insertions(+), 2 deletions(-) |
The port doesn't build on armv7. I can try to investigate, but will be busy for the next few days. The undeclared function looks like it's a NEON intrinsic. So probably some compiler option is missing or it tries to build aarch64 NEON code on armv7. [ 6% 18/283] /usr/local/libexec/ccache/cc -DOPENEXRCORE_EXPORTS -DOpenEXRCore_EXPORTS -I/wrkdirs/usr/ports/graphics/openexr/work/.build/src/lib/OpenEXRCore -I/wrkdirs/usr/ports/graphics/openexr/work/openexr-3.1.6/src/lib/OpenEXRCore -I/wrkdirs/usr/ports/graphics/openexr/work/.build/cmake -isystem /usr/local/include -isystem /usr/local/include/Imath -O2 -pipe -fstack-protector-strong -fno-strict-aliasing -O2 -pipe -fstack-protector-strong -fno-strict-aliasing -DNDEBUG -fPIC -fvisibility=hidden -MD -MT src/lib/OpenEXRCore/CMakeFiles/OpenEXRCore.dir/internal_zip.c.o -MF src/lib/OpenEXRCore/CMakeFiles/OpenEXRCore.dir/internal_zip.c.o.d -o src/lib/OpenEXRCore/CMakeFiles/OpenEXRCore.dir/internal_zip.c.o -c /wrkdirs/usr/ports/graphics/openexr/work/openexr-3.1.6/src/lib/OpenEXRCore/internal_zip.c FAILED: src/lib/OpenEXRCore/CMakeFiles/OpenEXRCore.dir/internal_zip.c.o /usr/local/libexec/ccache/cc -DOPENEXRCORE_EXPORTS -DOpenEXRCore_EXPORTS -I/wrkdirs/usr/ports/graphics/openexr/work/.build/src/lib/OpenEXRCore -I/wrkdirs/usr/ports/graphics/openexr/work/openexr-3.1.6/src/lib/OpenEXRCore -I/wrkdirs/usr/ports/graphics/openexr/work/.build/cmake -isystem /usr/local/include -isystem /usr/local/include/Imath -O2 -pipe -fstack-protector-strong -fno-strict-aliasing -O2 -pipe -fstack-protector-strong -fno-strict-aliasing -DNDEBUG -fPIC -fvisibility=hidden -MD -MT src/lib/OpenEXRCore/CMakeFiles/OpenEXRCore.dir/internal_zip.c.o -MF src/lib/OpenEXRCore/CMakeFiles/OpenEXRCore.dir/internal_zip.c.o.d -o src/lib/OpenEXRCore/CMakeFiles/OpenEXRCore.dir/internal_zip.c.o -c /wrkdirs/usr/ports/graphics/openexr/work/openexr-3.1.6/src/lib/OpenEXRCore/internal_zip.c /wrkdirs/usr/ports/graphics/openexr/work/openexr-3.1.6/src/lib/OpenEXRCore/internal_zip.c:118:17: warning: implicit declaration of function 'vqtbl1q_u8' is invalid in C99 [-Wimplicit-function-declaration] vPrev = vqtbl1q_u8 (d, shuffleMask); ^ /wrkdirs/usr/ports/graphics/openexr/work/openexr-3.1.6/src/lib/OpenEXRCore/internal_zip.c:118:15: error: assigning to 'uint8x16_t' (vector of 16 'uint8_t' values) from incompatible type 'int' vPrev = vqtbl1q_u8 (d, shuffleMask); ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~ /wrkdirs/usr/ports/graphics/openexr/work/openexr-3.1.6/src/lib/OpenEXRCore/internal_zip.c:190:25: warning: implicit declaration of function 'vzip1q_u8' is invalid in C99 [-Wimplicit-function-declaration] uint8x16_t lo = vzip1q_u8 (a, b); ^ /wrkdirs/usr/ports/graphics/openexr/work/openexr-3.1.6/src/lib/OpenEXRCore/internal_zip.c:190:20: error: initializing 'uint8x16_t' (vector of 16 'uint8_t' values) with an expression of incompatible type 'int' uint8x16_t lo = vzip1q_u8 (a, b); ^ ~~~~~~~~~~~~~~~~ /wrkdirs/usr/ports/graphics/openexr/work/openexr-3.1.6/src/lib/OpenEXRCore/internal_zip.c:191:25: warning: implicit declaration of function 'vzip2q_u8' is invalid in C99 [-Wimplicit-function-declaration] uint8x16_t hi = vzip2q_u8 (a, b); ^ /wrkdirs/usr/ports/graphics/openexr/work/openexr-3.1.6/src/lib/OpenEXRCore/internal_zip.c:191:20: error: initializing 'uint8x16_t' (vector of 16 'uint8_t' values) with an expression of incompatible type 'int' uint8x16_t hi = vzip2q_u8 (a, b); ^ ~~~~~~~~~~~~~~~~ 3 warnings and 3 errors generated.