Hi apr people, Alle domenica 22 gennaio 2012, Pino Toscano ha scritto: > Package: apr > Version: 1.4.5-1.1 > Severity: important > Tags: patch > User: email@example.com > Usertags: hurd > > Hi, > > at the moment, the shared memory and file locking parts of apr don't > work correctly on GNU/Hurd. > > The problem is that there are some methods for shared memory and file > locking that are not implemented on Hurd, yet, causing failures in > the test suite (which is currently disabled for Hurd). Luckly, there > are few methods which actually work, so we can temporarly force > their usage. > > Furthermore, there is a small issue in the way arch-specific hints > are set, making the Hurd ones not used. > > * hurd_fix-hints.diff > This patch, which could be sent upstream, fixes the set of > predefined hints for Hurd: usually the $host string is > "i686-unknown-gnu0.3" or "i486-pc-gnu", so the "GNU" string won't > match anywhere; furthermore, since the proposed string could be too > broad (potentially matching also the kFreeBSD string, e.g. > "i686-unknown-kfreebsd8.2-gnu"), it has been moved as very last > check. > > * hurd_shm_flock.diff > This patch disables all the methods for shared memory and file > locking found by the configure checks, forcing the only working > ones. This patch should stay Debian-specific, as it is not that > suitable for upstream inclusion. (As you can guess, ideally all of > those non-working methods should be ideally fixed in Hurd...) > > * debian_enabletests.diff > This patch reenables the test suite on Hurd, which should pass now > with the above patches. Any news about the above patches? > (Note that testsockets will still fail, but > a testsockets+testsock run failure is ignored; the bug is a wrong > errno code set in hurd's version of socket() in glibc, and it has > been reported already.) (this has been fixed in glibc in the meanwhile.) Thanks, -- Pino Toscano
Description: This is a digitally signed message part.