MySQL 4.0.16 and 4.0.17 fail to complete the ./configure when using a 2.6 kernel with a glibc with NPTL enabled. The problem is that the configure script does some sort of literal check for LinuxThreads, which it doesn't find. The solution is to force MySQL to skip the test and just link to the pthread library, see the patch below (should apply to 4.0.16 as well): --- mysql-4.0.17.ebuild-orig 2004-02-17 11:43:25.287212596 +0100 +++ mysql-4.0.17.ebuild 2004-02-17 11:46:21.482117401 +0100 @@ -112,6 +112,10 @@ || myconf="${myconf} --without-berkeley-db" fi + # When using NPTL instead of LinuxThreads, MySQL configure complains + # about not finding LinuxThreads + use nptl && myconf="${myconf} --with-pthreads" + #readline pair reads backwards on purpose, DONT change it around, Ok? # this is because it refers to the building of a bundled readline # versus the system copy
since 4.0.18 is stable, is this still an issue ?
It does not seem to be a problem anymore, but I've taken a look at the patch that is used to work around the detection and IMHO it's quite a dirty hack and my patch to the ebuild is nicer (although still a hack, only less dirty :).
Bump. Time to close this one? I do have MySQL 4.0.18 up and running nicely with nptl enabled in glibc.
This was fixed some time ago in the ebuilds, but also note that the current patch to configure.in is also required to make 4.1.4 (not in portage) compile, based on a manual build I did.
close this as nobody has complained