| Summary: | Host contamination in glibc-locales | ||
|---|---|---|---|
| Product: | [Build System, Metadata & Runtime] OE-Core | Reporter: | Ross Burton <ross.burton> |
| Component: | core | Assignee: | Juro Bystricky <juro.bystricky> |
| Status: | RESOLVED DUPLICATE | QA Contact: | |
| Severity: | normal | ||
| Priority: | Medium | CC: | joshuagloe, juro.bystricky, li.zhou, meta.mr.watcher, meta.watcher, randy.macleod, seebs |
| Version: | 1.9 | ||
| Target Milestone: | 2.3 | ||
| Hardware: | All | ||
| OS: | Multiple | ||
| Whiteboard: | |||
| OS type for building Yocto: | --- | Type of Regression: | --- |
| Verified: | Documentation change: | No (bug/feature does not impact docs) | |
|
Description
Ross Burton
2015-10-06 11:01:14 UTC
This was likely caused by a bug in pseudo, which is now fixed in oe-core 4bd658f7dae76ae8790fd1dfdd89fa58a456a0a4. Replicated with current master (poky a1a0ce1a0064e29159555011ebd667192132e0ce) with a clean tmp and no sstate. Considering this is random and doesn't always hit, I'd still consider pseudo to be a suspect here. Peter, what's a good way to start investigating this from the pseudo side? Inspect the log or sqlite database for clues? More data: WARNING: glibc-locale-2.24-r0 do_package_qa: QA Issue: glibc-locale: /glibc-binary-localedata-fr-lu/usr/lib/locale/fr_LU/LC_MEASUREMENT is owned by uid 1000, which is the same as the user running bitbake. This may be due to host contamination [host-user-contaminated] $ dpkg -c glibc-binary-localedata-fr-lu_* drwxrwxrwx root/root 0 2016-06-14 14:26 ./ drwxr-xr-x root/root 0 2016-06-14 14:23 ./usr/ drwxr-xr-x root/root 0 2016-06-14 14:23 ./usr/lib/ drwxr-xr-x root/root 0 2016-06-14 14:23 ./usr/lib/locale/ drwxr-xr-x root/root 0 2016-06-14 14:23 ./usr/lib/locale/fr_LU/ drwxr-xr-x root/root 0 2016-06-14 14:23 ./usr/lib/locale/fr_LU/LC_MESSAGES/ -rw-r--r-- root/root 60 2016-06-14 14:22 ./usr/lib/locale/fr_LU/LC_MESSAGES/SYS_LC_MESSAGES -rw-r--r-- root/root 1243766 2016-06-14 14:22 ./usr/lib/locale/fr_LU/LC_COLLATE -rw-r--r-- root/root 2354 2016-06-14 14:22 ./usr/lib/locale/fr_LU/LC_TIME -rw-r--r-- root/root 156 2016-06-14 14:22 ./usr/lib/locale/fr_LU/LC_ADDRESS -rw-r--r-- root/root 62 2016-06-14 14:22 ./usr/lib/locale/fr_LU/LC_NAME -rw-r--r-- root/root 50 2016-06-14 14:22 ./usr/lib/locale/fr_LU/LC_TELEPHONE -rw-r--r-- root/root 348 2016-06-14 14:22 ./usr/lib/locale/fr_LU/LC_IDENTIFICATION -rw-r--r-- root/root 34 2016-06-14 14:22 ./usr/lib/locale/fr_LU/LC_PAPER -rw-r--r-- root/root 54 2016-06-14 14:22 ./usr/lib/locale/fr_LU/LC_NUMERIC -rw-r--r-- 1000/pulse 23 2016-06-14 14:22 ./usr/lib/locale/fr_LU/LC_MEASUREMENT -rw-r--r-- root/root 317872 2016-06-14 14:22 ./usr/lib/locale/fr_LU/LC_CTYPE -rw-r--r-- root/root 294 2016-06-14 14:22 ./usr/lib/locale/fr_LU/LC_MONETARY Peter, the file that gets the wrong ownership appears to change randomly so I'm fairly convinced that this is due to pseudo losing track of files somehow. I met such issue occasionally too, and hard to reproduce. I notice there is pseudo upgrade recently. Does it work on this issue? Thanks. This could be related to the hardlinking behavior previously discussed. Are any of these files hard links to files that aren't tracked in pseudo? @Seebs:
Thanks for your answer. Is there any link to the discussion you mentioned? Per my understanding to your answer, those files shouldn't be hard links. The files that have been recorded are listed as below:
/glibc-binary-localedata-es-es/usr/lib64/locale/es_ES/LC_MEASUREMENT
/glibc-binary-localedata-pap-aw/usr/lib/locale/pap_AW/LC_TIME
/glibc-binary-localedata-pa-pk/usr/lib64/locale/pa_PK/LC_MESSAGES/SYS_LC_MESSAGES
/glibc-binary-localedata-ar-kw/usr/lib64/locale/ar_KW/LC_CTYPE
They are quite similar with this original issue </glibc-binary-localedata-da-dk.iso-8859-1/usr/lib/locale/da_DK.ISO-8859-1/LC_PAPER>.
I think there are many similar locale files in those locale packages. And we just occasionally get the wrong uid with a very low frequency (each time with a different file).
After we integrate http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/meta?id=577ff4e3dd4efe3b129a10eabedf620cfb7a2fa8 , the frequency of this issue has certainly decreased from 5-10 times/ month to < 1 time / month. Since the commit was just in our code from mid September, we only have done limit observation. We'll continue to watch for additional errors in the next month. And after a month of running 100s of builds/day, we still have NOT seen another failure so this defect can be closes @seebs. (In reply to comment #10) > And after a month of running 100s of builds/day, we still have NOT seen > another failure so this defect can be closes @seebs. This was observed again, I actually filed a new bug report, unaware of this one. https://bugzilla.yoctoproject.org/show_bug.cgi?id=11299 So I close this one as a duplicate. *** This bug has been marked as a duplicate of bug 11299 *** |