Summary: | www-client/firefox-16.0.1 - .../work/mozilla-release/memory/jemalloc/src/src/jemalloc.c:1736:1: error: corrupted profile info: edge from 0 to 2 exceeds maximal count | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Wei-Chun Chung <moneycat.tw> |
Component: | [OLD] Unspecified | Assignee: | Mozilla Gentoo Team <mozilla> |
Status: | RESOLVED WONTFIX | ||
Severity: | normal | CC: | asturm, betelgeuse, coatzin, elettronik, f_philipp, kjackie, main.haarp, mattsch, rauchwolke, vityokster |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | x86 | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
build.log.tar.bz2
pgo build patch |
Description
Wei-Chun Chung
2012-10-22 09:48:03 UTC
Created attachment 327142 [details]
build.log.tar.bz2
The last lines of the file:
/var/tmp/portage/www-client/firefox-16.0.1/work/mozilla-release/memory/jemalloc/src/src/jemalloc.c: In function 'tcache_event':
/var/tmp/portage/www-client/firefox-16.0.1/work/mozilla-release/memory/jemalloc/src/src/jemalloc.c:1736:1: error: corrupted profile info: edge from 0 to 2 exceeds maximal count
/var/tmp/portage/www-client/firefox-16.0.1/work/mozilla-release/memory/jemalloc/src/src/jemalloc.c: In function 'arena_dalloc':
/var/tmp/portage/www-client/firefox-16.0.1/work/mozilla-release/memory/jemalloc/src/src/jemalloc.c:1736:1: note: correcting inconsistent profile data
make[5]: *** [jemalloc.o] Error 1
make[5]: *** Waiting for unfinished jobs....
make[5]: Leaving directory `/var/tmp/portage/www-client/firefox-16.0.1/work/mozilla-release/obj-i686-pc-linux-gnu/memory/jemalloc'
make[4]: *** [libs_tier_base] Error 2
make[4]: Leaving directory `/var/tmp/portage/www-client/firefox-16.0.1/work/mozilla-release/obj-i686-pc-linux-gnu'
make[3]: *** [tier_base] Error 2
make[3]: Leaving directory `/var/tmp/portage/www-client/firefox-16.0.1/work/mozilla-release/obj-i686-pc-linux-gnu'
make[2]: *** [default] Error 2
make[2]: Leaving directory `/var/tmp/portage/www-client/firefox-16.0.1/work/mozilla-release/obj-i686-pc-linux-gnu'
make[1]: *** [realbuild] Error 2
make[1]: Leaving directory `/var/tmp/portage/www-client/firefox-16.0.1/work/mozilla-release'
make: *** [profiledbuild] Error 2
emake failed
* ERROR: www-client/firefox-16.0.1 failed (compile phase):
* virtualmake: the emake failed.
*
* Call stack:
* ebuild.sh, line 85: Called src_compile
* environment, line 6844: Called Xemake '-f' 'client.mk' 'profiledbuild'
* environment, line 673: Called virtualmake '-f' 'client.mk' 'profiledbuild'
* environment, line 7605: Called die
* The specific snippet of code:
* [[ ${retval} -ne 0 ]] && die "${FUNCNAME}: the ${VIRTUALX_COMMAND} failed.";
*
* If you need support, post the output of `emerge --info '=www-client/firefox-16.0.1'`,
* the complete build log and the output of `emerge -pqv '=www-client/firefox-16.0.1'`.
* The complete build log is located at '/var/tmp/portage/www-client/firefox-16.0.1/temp/build.log'.
* The ebuild environment file is located at '/var/tmp/portage/www-client/firefox-16.0.1/temp/environment'.
* Working directory: '/var/tmp/portage/www-client/firefox-16.0.1/work/mozilla-release'
* S: '/var/tmp/portage/www-client/firefox-16.0.1/work/mozilla-release'
Comment on attachment 327142 [details]
build.log.tar.bz2
There is no need to store a single file in a tar archive and then compress it - simply compressing a single file will do.
same error here Compile with -pgo, I am gonna be masking the useflag by weeks end due to all the failures. Same error here on amd64. I can reproduce this bug only with gcc-4.5.4 on amd64 not with gcc-4.6.3 on amd64. Also confirmed here on AMD64. Had no problems with firefox-15.0.1 (previous version). The build works for me when removing the pgo useflag: # USE="-pgo" emerge -1v =www-client/firefox-16.0.1 I've googled and apparently there were previous build errors like this: https://bugzilla.mozilla.org/show_bug.cgi?id=418866 (This also gave me the idea to try disable PGO.) Hope this helps, as a temporary work-around. Created attachment 327250 [details, diff]
pgo build patch
If it works with gcc-4.6, pgo should only be masked for gcc-4.5 and lower.
How about that?
I don't think mask firefox-16 of gcc version less than 4.6 is a good idea because current stable (for most user) is 4.5.4. (In reply to comment #9) > I don't think mask firefox-16 of gcc version less than 4.6 is a good idea > because current stable (for most user) is 4.5.4. Only with the pgo useflag set it would be masked, so users would have to remove this useflag (if set), hence the build would succeed. FYI, I emerged gcc-4.6.3 for the purpose of building firefox-16 with USE="pgo" and it was successful. The rest of the system is build with gcc-4.5.4 and I also returned to 4.5.4 after the successful merge. There is the same issue with 16.0.2. Compiling with USE="pgo" succeeds only with gcc-4.6 (tested with sys-devel/gcc-4.6.3, cmdline: emerge firefox). Compiling with gcc-4.5 works without the pgo useflag (tested with sys-devel/gcc-4.5.4, cmdline: USE="-pgo" ebuild firefox-16.0.2.ebuild compile). Confirmed on x86 with gcc-4.5.4 and +pgo flag. No problem without pgo I found that upstream maintainers know about this problem but they don't have any solution: https://bugzilla.mozilla.org/show_bug.cgi?id=419470 Can we use system jemalloc instead of mozilla's copy? *** Bug 442144 has been marked as a duplicate of this bug. *** Seeing problem for firefox 17 too. I have added 17.0-r2 to the mozilla overlay please test it and report back. Same error with r2 than originally reported The 17.0-r2 still failed to compile with pgo and gcc-4.5, maybe the pgo flag should only be enabled with gcc-4.6 or higher. On my system, pgo is now masked. I just can’t figure out how (in which file) it is masked. (In reply to comment #19) > On my system, pgo is now masked. I just can’t figure out how (in which file) > it is masked. You can unmask pgo just adding -pgo at /etc/portage/profile/use.mask file cat /etc/portage/profile/use.mask -pgo After upgrading to gcc-4.6.3 on x86 stable, the compiling of firefox-16/17 with pgo works normally. firefox-18.0[+pgo] compiled and working fine on amd64 with GCC 4.6.3. One weird issue though with firefox-18 - it is missing my locale settings: whole interface is in English. Not sure if this is the issue of pgo USE flag or firefox-18 itself. Be warned. (In reply to comment #22) > firefox-18.0[+pgo] compiled and working fine on amd64 with GCC 4.6.3. > > One weird issue though with firefox-18 - it is missing my locale settings: > whole interface is in English. Not sure if this is the issue of pgo USE flag > or firefox-18 itself. Be warned. I have this issue too. What I did to fix this was disable and enable again the language and restart firefox. (In reply to comment #23) > What I did to fix this was disable and enable again > the language and restart firefox. Many thanks, it helped me too! Now that gcc-4.6.3 is stable, I guess pgo can be unmasked again? (In reply to comment #25) > Now that gcc-4.6.3 is stable, I guess pgo can be unmasked again? Just cause one issue is addressed by a compiler update does not warrant unmasking pgo useflag again. There are many bugs that need to be addressed and we the mozilla team need to be able to maintain it on a much larger scale of users. Sorry, I got the impression that all of this bug was solved by the gcc-4.6.3 upgrade. Anyway, firefox-17.0.2 also built fine with gcc-4.7.2 here. (In reply to comment #27) > Sorry, I got the impression that all of this bug was solved by the gcc-4.6.3 > upgrade. I compiled firefox-18.0 using gcc-4.6.3 on my AMD64 box with USE="pgo" and it does work, like with previous versions of firefox and pgo. It also runs without any errors. The only issue is the language thing as described in comment #22 by Coacher, but I seriously doubt that this is related to pgo. As stated in use.mask pgo is unsupported. |