On Wed, Nov 13, 2019 at 11:38:55AM +0100, Svante Signell wrote:
> Source: e2fsprogs
> Version: 1.45.4-1
> Severity: important
> Tags: patch
> Usertags: hurd
> User: debian-h...@lists.debian.org
>
> Hello,
>
> The latest version of e2fsprogs in sid, 1.45.4-1, FTBFS on GNU/Hurd due to two
> reasons:
> 1) <sys/mount.h> is not available.
> 2) PATH_MAX is not defined.
>
> 1) is fixed with the attached patch, configure.ac.diff where a variable MOUNT
> is
> introduced and checked for.
That's not the correct patch. The right fix is to drop the test for
sys/mount in AX_CHECK_MOUNT_OPT in acinclude.m4. If sys/mount doesn't
exist, this will be interpreted as the system not supporting nosuid
and nodev. (Which is fine, since as I recall HURD doesn't support
fuse, and HAVE_MOUNT_{NOSUID,NODEV} is only used for fuse2fs.)
> 2) is fixed by lib_ext2fs_dirhash.c.diff, which simply defines PATH_MAX, if
> not
> defined. Instead of using PATH_MAX this could probably be solved more
> elegantly
> by dynamic allocation of the needed size of buff, but grepping for PATH_MAX in
> the code shows plenty of occurrences already, solved in various ways, so I did
> not find the motivation for doing that.
Thanks, that's look good. I'll apply that for the next release.
> Additionally, in order to test the packages on a GNU/Linux system, not using
> systemd, patches debian_control.diff and debian_e2fsprogs.install were needed.
Yeah, but that's not something I can apply, because we need systemd to
be present on the build boxes so we can properly create the e2fsprogs
debs to support systemd systems --- and the only way to enforce this
is with a build dependency.
> Regarding the tests for GNU/Hurd they have to be disabled for now. Plenty of
> failing tests are not applicable to Hurd. Patches for the failing tests are in
> the works, the list is now 38 entries with some more to fix. That patch is
> pending and will be reported in a separate bug report.
Please ensure that the any changes to the tests do not cause them to
break on Linux, and if possible it would be nice if you could check to
make sure it doesn't cause test regressions on FreeBSD as well.
Thanks!
- Ted
