I need to find an ia64 machine to investigate this build failure:
https://buildd.debian.org/status/fetch.php?pkg=e2fsprogs&arch=ia64&ver=1.45.3-1&stamp=1563898609&raw=0
I couldn't find any in the Debian machines list; are there any
available?
Failing that, if someone could send me the file
tests/m_hugefile.failed from the build tree after running "make check"
on an ia64 machine, that would be really helpful.
Hi Ted!
On 7/23/19 11:22 PM, Theodore Y. Ts'o wrote:
I need to find an ia64 machine to investigate this build failure:
https://buildd.debian.org/status/fetch.php?pkg=e2fsprogs&arch=ia64&ver=1.45.3-1&stamp=1563898609&raw=0
Thanks for access. It looks like it works just fine on the machine
you gave me access. The test in question, m_hugefile, creates a 4T
file system. Assuming the file system on the build host for /tmp
supports sparse files, it requires 1.1G of space. If it doesn't
support sparse files, then it needs 4TB of space:
% mke2fs -t ext4 -Fq /tmp/foo.img 4T
% ls -slh /tmp/foo.img
1.1G -rw-r--r-- 1 tytso tytso 4.0T Jul 25 21:18 /tmp/foo.img
Is there is a possibility that the problem is the lack of space on
/tmp on the ia-64 build server?
Is there is a possibility that the problem is the lack of space on
/tmp on the ia-64 build server?
/tmp is part of the root file system which has 6.2G of free space,
it might have been less during the build. I'll see if I can do some
clean up.
Thanks. I just pushed a new release of e2fsprogs 1.45.3-3. That
should fix the other build failures on x32 and kfreebsd-*. If the
clean up fixes the build, then we should be 100% clean on all of the
Debian ports builds. Otherwise, I can look at skipping m_hugefile
when running "make check" out of debian/rules.
[1] https://git.kernel.org/pub/scm/fs/ext2/e2fsprogs.git/commit/?id=7f4c3bb1200eb727894ddfad6a99167bdf2eb78f
[2] https://salsa.debian.org/libreoffice-team/libreoffice/libreoffice/blob/master/rules#L216
Thanks for the updated package. Unfortunately, you introduced a minor regression with the override for the dh_auto_test target by disabling
the "nocheck" mechanism [1].
This means, it's no longer possible to disable the testsuite via environment flags and hence it will be run unconditionally which is a problem on architectures
like m68k and sh4 where we are currently building the packages with qemu.
Could you add a nocheck condition around the "make check" call like here [2]?
Sysop: | Keyop |
---|---|
Location: | Huddersfield, West Yorkshire, UK |
Users: | 292 |
Nodes: | 16 (2 / 14) |
Uptime: | 202:24:50 |
Calls: | 6,617 |
Calls today: | 1 |
Files: | 12,168 |
Messages: | 5,316,354 |