On 2015-03-05 12:09, Ben Hutchings wrote: > On Sat, 2015-02-28 at 21:02 +0000, Ben Hutchings wrote: > > I've started preparing an update to eglibc, fixing the many open > > security issues of lower severity than those fixed recently. > > > > I didn't yet had time to test it properly, so I've uploaded my work to > > <https://people.debian.org/~benh/packages/squeeze-lts/> so others can > > test further and maybe complete the update. The version number there is > > 2.11.3-4+deb6u5~benh.1 but the real update should be 2.11.3-4+deb6u5. > > > > There don't appear to be any regressions in compiler warnings, and > > nothing went obviously wrong when I installed the new libraries in a VM. > > > > However, for the current version, 2.11.3-4+deb6u4, the build log shows 1 > > unexpected test suite failure for each of the library configurations: > > > > Encountered regressions that don't match expected failures: > > bug-regex32.out, Error 1 > > > > while for my version, 2.11.3-4+deb6u5~benh.1, there were several more: > > > > [i386] > > Encountered regressions that don't match expected failures: > > bug-regex32.out, Error 1 > > tst-cancel4.out, Error 1 > > tst-cancel5.out, Error 1 > > tst-nice.out, Error 1 > [...] > > I now understand all the regressions: > > - tst-nice failed because I run pbuilder with 'nice -19' and the test > process then can't change its niceness. It passes without this. > - tst-cancel4 (and the three variants compiled in different ways) fails > as the minimum kernel socket buffer size is now larger than it > assumes. I cherry-picked a fix from glibc git and it now passes. Sorry for not answering earlier. I confirm that your analysis is correct. Unfortunately a lot of tests are failing or passing depending on the kernel version. -- Aurelien Jarno GPG: 4096R/1DDD8C9B aurelien@aurel32.net http://www.aurel32.net
Attachment:
signature.asc
Description: Digital signature