| Summary: | dev-util/ccache breaks net-analyzer/wireshark-1.0.2, app-office/openoffice build | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Eugene Volchek <eugene.volchek> |
| Component: | Current packages | Assignee: | Gentoo Toolchain Maintainers <toolchain> |
| Status: | RESOLVED NEEDINFO | ||
| Severity: | normal | CC: | netmon, pacho, pva, robbat2, suka |
| Priority: | High | ||
| Version: | unspecified | ||
| Hardware: | x86 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: | Entire build log | ||
|
Description
Eugene Volchek
2008-09-09 14:47:49 UTC
Could you try rebuilding with MAKEOPTS=-j1 please? And do post more of the build output, because the lines where .libs/msg_dreg.o is built are omitted. Created attachment 165033 [details]
Entire build log
The problem is in ccache. I tried to turn ccache off and wireshark emerged without a problem. When I turn ccache back on, the problem comes back. So the issue turns into ccache problem. Do you still reproduce this bug? Have you changed compiler recently? @toolchain: do you have any ideas why this could happen? As I undrestand compiler change could cause this and if my suspect is correct I think it's good idea to clear ccache on each gcc change. What do you think? I tried to reemerge wireshark with the other CCACHE_DIR and succeeded. It is strange because my system is about a week old and I didn't change gcc and it was initial wireshark emerge. Anyway, my problem has been resolved but ccache potential problem may still remain. Eugene that question was is it possible to reproduce bug you your machine? I think it's good idea to investigate this problem further and ask help on ccache development list, but if you unable to reproduce bug now we'll have to close it as-is. Yes, I can. I did'n clear the cache but change CCACHE_DIR instead. I can reproduce it using old CCACHE_DIR. you shouldn't need to clear the ccache when you change compilers. ccache hashes the compiler executable mtime so it won't use cached data created with another compiler version. worst case situation you have data in cache that can't be used, but it automatically gets wiped as you run out of space. Thank you Ryan. But then can you suggest further steps how to debug this problem? no idea unfortunately. i imagine a corrupted ccache dir could cause problems as once ccache things its cached files are OK, it wont recheck things I get the same problem in one of my x86-64 systems when merging openoffice-3.0 with ccache enabled (it success with it disabled):
echo unxlngx6.pro/slo/regimpl.o unxlngx6.pro/slo/regkey.o unxlngx6.pro/slo/registry.o unxlngx6.pro/slo/keyimpl.o unxlngx6.pro/slo/reflread.o unxlngx6.pro/slo/reflwrit.o | xargs -n1 > ../unxlngx6.pro/slb/regcpp.lib
nm: ../unxlngx6.pro/slo/reflwrit.o: File format not recognized
dmake: Error code 1, while making '../unxlngx6.pro/slb/regcpp.lib'
dmake: '../unxlngx6.pro/slb/regcpp.lib' removed.
Running processes: 0
1 module(s):
registry
need(s) to be rebuilt
Reason(s):
ERROR: error 65280 occurred while making /var/tmp/portage/app-office/openoffice-3.0.0/work/ooo/build/ooo300-m9/registry/source
Attention: if you build and deliver the above module(s) you may prolongue your the build issuing command "build --from registry"
rmdir /tmp/3978
make: *** [stamp/build] Error 1
Can this be handled in this bug report or should I open another one? Thanks
I don't think you need to open new bug. I just CC openoffice maintainer here, so he'll be aware about this bug. At this moment we have no solution for this bug besides idea that this bug is cause by cache corruption. for somebody that only changed their CCACHE_DIR, could you please find the files with the matching path and name where one caused the problem, and the other didn't, and attach both .o files? Robin asked for some info and never got it...I'm putting this to needinfo. |
