In file included from source/test/testbench.cpp:29: In file included from source/test/pixelharness.h:28: source/test/testharness.h:74:24: error: static declaration of '__rdtsc' follows non-static declaration static inline uint32_t __rdtsc(void) ^ source/test/testharness.h:74:24: note: previous implicit declaration is here build log: http://sprunge.us/TFVf
Probably a Clang bug. $ cat a.c void __rdtsc(void) {} $ cc -fno-builtin -c a.c a.c:1:6: error: conflicting types for '__rdtsc' void __rdtsc(void) {} ^ a.c:1:6: note: '__rdtsc' is a builtin with type 'unsigned long long (void)' a.c:1:6: error: definition of builtin function '__rdtsc' void __rdtsc(void) {} ^ 2 errors generated.
I'm curious, where the reporter obtained clang-4.0 to begin with... I don't see such a thing under lang/ at all -- the lang/clang-devel installs 5.x
A commit references this bug: Author: mi Date: Tue Jan 17 20:43:11 UTC 2017 New revision: 431774 URL: https://svnweb.freebsd.org/changeset/ports/431774 Log: Upgrade from 1.9 to 2.2. Spurred into action by PR: 216162 Changes: head/multimedia/x265/Makefile head/multimedia/x265/distinfo head/multimedia/x265/pkg-plist
(In reply to Mikhail Teterin from comment #2) > where the reporter obtained clang-4.0 to begin with... https://svnweb.freebsd.org/base/projects/clang400-import/?view=log $ poudriere jail -cj clang40-i386 -a i386 -v projects/clang400-import -m svn (In reply to commit-hook from comment #3) > Upgrade from 1.9 to 2.2. Have you tested consumers aren't broken? At least bump their PORTREVISION since SOVERSION has changed. https://abi-laboratory.pro/tracker/timeline/x265/
Yeah, Brooks will probably setup a clang40 port, once the upstream branch gets its first release candidate. But does x265 2.2 fix the build problem reported by Jan?
(In reply to Jan Beich (mail not working) from comment #4) > Have you tested consumers aren't broken? This reminds me of the old Soviet joke about a bureaucrat demanding, an applicant proves his "single" status by presenting an affidavit from ALL WOMEN stating, she is not his wife. I tested with the vendor's own test-harness. > But does x265 2.2 fix the build problem reported by Jan? No, most likely not. I'm still working on that. May need to wait for lang/clang4...
Created attachment 179007 [details] Possible fix Jan could you, please, try again with this file dropped into files/ subdirectory? Thanks!
Comment on attachment 179007 [details] Possible fix Builds fine with the patch: http://sprunge.us/XgWH Also tested locally on amd64 (haven't created poudriere jail yet).
A commit references this bug: Author: mi Date: Wed Jan 18 00:31:11 UTC 2017 New revision: 431776 URL: https://svnweb.freebsd.org/changeset/ports/431776 Log: Add a patch to avoid providing our own implementation of __rtdsc when built with clang version 4 or higher (which has its own). PR: 216162 Changes: head/multimedia/x265/files/patch-rdtsc
The likely fix has been committed. Tested with clang-4 by jbeich (the reporter) and clang-5 by myself (the maintainer). Please, reopen if anything is still broken. Thanks!
A commit references this bug: Author: jbeich Date: Thu Jan 19 05:39:23 UTC 2017 New revision: 431851 URL: https://svnweb.freebsd.org/changeset/ports/431851 Log: MFH: r431776 by mi Add a patch to avoid providing our own implementation of __rtdsc when built with clang version 4 or higher (which has its own). PR: 216162 Approved by: ports-secteam blanket Changes: _U branches/2017Q1/ branches/2017Q1/multimedia/x265/files/patch-rdtsc