PR: 244625
Details
Diff Detail
- Repository
- rP FreeBSD ports repository
- Lint
Lint Not Applicable - Unit
Tests Not Applicable
Event Timeline
I just did a poudriere testport and there is a plist issue: http://pkg.awarnach.mathstat.dal.ca/data/12amd64-default/2020-03-06_12h24m58s/logs/errors/chromium-80.0.3987.132.log
Do you mind if I update the diff here?
Hm, I did a check-plist (non-poudriere). Let's see what your log shows. (That logfile is a massive 144MB, FWIW.) The entire relevant part seems to be these 265 bytes:
====> Checking for pkg-plist issues (check-plist) ===> Parsing plist ===> Checking for items in STAGEDIR missing from pkg-plist ===> Checking for items in pkg-plist which are not in STAGEDIR Error: Missing: libdata/ldconfig/chromium ===> Error: Plist issues found.
So in my version of ports, www/chromium installs that file and check-plist warns if you *don't* include it. I wonder why it's different in your environment?
$ ls -l work/stage/usr/local/libdata/ldconfig/chromium -rw-r--r-- 1 conrad wheel 26 Mar 5 21:42 work/stage/usr/local/libdata/ldconfig/chromium
# USE_LDCONFIG - If set to "yes", this adds ${PREFIX}/lib to the list of # directories to be searched for shared libraries. # Otherwise, this is a list of directories to be added to that # list. The directory names are written to # ${LOCALBASE}/libdata/ldconfig/${PKGBASE} which is then # used by the ldconfig startup script.
And the port sets USE_LDCONFIG=yes...
I'm happy to remove it if that's the correct thing to do, but I'd like to understand the discrepancy.
Hm, is LOCALBASE != PREFIX in your environment? Maybe the right plist entry would be explicit /usr/local/libdata/...?
This is a vanilla 12.1 amd64 jail with default options, so AFAIK the same jail that the official packages will be build with. I just updated the i386 jail, so I'll test there as well. All variables are defaults, including LOCALBASE.
Drop ldconfig line from pkg-plist, which seems to be the issue Joseph observes
in Poudriere.
The build fails on 11.3 i386.
In file included from ../../base/profiler/stack_copier_suspend.cc:7: In file included from ../../base/profiler/metadata_recorder.h:14: In file included from ../../base/synchronization/lock.h:11: ../../base/synchronization/lock_impl.h:71:12: warning: releasing mutex 'native_handle_' that was not held [-Wthread-safety-analysis] int rv = pthread_mutex_unlock(&native_handle_); ^ ../../base/profiler/stack_copier_suspend.cc:51:14: error: use of undeclared identifier 'RegisterContextStackPointer' bottom = RegisterContextStackPointer(thread_context); ^ 1 warning and 1 error generated.
Here is the full build log: http://pkg.awarnach.mathstat.dal.ca/data/11i386-default/2020-03-06_23h58m22s/logs/errors/chromium-80.0.3987.132.log
The build also fails on 12.1 i386 with the same error.
http://pkg.awarnach.mathstat.dal.ca/data/12i386-default/2020-03-06_20h50m52s/logs/errors/chromium-80.0.3987.132.log
No, the work was done prior to r527876.
The i386 thing (at least that specific issue) can be fixed, but I don't know what else lurks.
- Rebase over r527876
- Add x86 bits to failed section (unclear what else will fail)
- Add preliminary (sketch, definitely broken) aarch64 support to same section. Someone more knowledgeable of aarch64 can fill in how to find the framepointer and programcounter.
www/chromium/files/patch-base_profiler_register__context.h | ||
---|---|---|
32–57 ↗ | (On Diff #69291) | I can confirm these new portions do not regress the amd64 build. |
Your i386 fixes look good. http://pkg.awarnach.mathstat.dal.ca/data/12i386-default/2020-03-07_13h52m51s/logs/chromium-80.0.3987.132.log
I'll test with your latest patch. Hopefully it will go a bit quicker with the hot ccache cache.