[Top] [All Lists]

Re: xfsprogs-3.1.11 pre-release please test!

To: xfs@xxxxxxxxxxx
Subject: Re: xfsprogs-3.1.11 pre-release please test!
From: "Michael L. Semon" <mlsemon35@xxxxxxxxx>
Date: Thu, 02 May 2013 09:34:42 -0400
Delivered-to: xfs@xxxxxxxxxxx
Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=x-received:message-id:date:from:user-agent:mime-version:to:subject :references:in-reply-to:content-type:content-transfer-encoding; bh=nk1wVdGn1NZeHmSj8czVQHAOWz8loII0cPAzd+rPEMU=; b=yWSBmqnkf7SiYuJk6CVQR5zy+OZySuKilJO8SBK/drP8nb2g4uq5wVXJsVCICr+Lzz qcVkt306wtC+9E5MKvPShI8BGYFlPD/OkP2czf3noygni/OVuT+Gx7ziKA5atbuhKOYi TA1dEsZB6W9QhYFfanA45fyYuNB+QSqNIf9s6iyi3Fbw3G9qV4PWWBvmH3amWPoVpxtg sK3B1TA4cYtfnoQVawJzCbZq+BczawhObF73MK0WyoFJW/MWFqML5p50Np1LSEUB8Lru BQuQPqEwtDxqYrNGtnfTkKtjMNnCRqdPMgclwQCMp5GgLHUC7leYLKg03++fk0LREX9P OUJg==
In-reply-to: <51817C88.4060208@xxxxxxx>
References: <51817712.6000807@xxxxxxx> <51817C88.4060208@xxxxxxx>
User-agent: Mozilla/5.0 (X11; Linux i686; rv:17.0) Gecko/20130328 Thunderbird/17.0.5
On 05/01/2013 04:35 PM, Mark Tinguely wrote:
On 05/01/13 15:12, Rich Johnston wrote:
Hi Folks,

Here are the changes for this release:

xfsprogs-3.1.11 (30 April 2013)

- Support for relative paths in xfs_quota thanks to Satoru Takeuchi.
- mkfs.xfs will always go into multidisk mode when filesystem
geometry is specified on the command line.
- Document all commands in xfs_io.
- Remove setfl command from xfs_io.
- xfs_metadump will obfuscate symlinks by path component.
- mkfs.xfs no longer accepts geometry settings smaller than the
physical sector size.
- xfs_logprint now supports multiply-logged inode fields and
handles continued inode transactions correctly.
- kill XLOG_SET
- Update release scripts to use git archive to address a
missing source file reported by Arkadiusz Mi?kiewicz
- Fix a build error with -Werror=format-security, reported
by Arkadiusz Mi?kiewicz
- mkfs.xfs no longer attempts to discard when -N option is used.
- Update 'make deb' to use tarball
- Sync up with log reservation changes in the kernel.
- Fix possible unallocated memory access in fiemap.
- Guard against string overflow in path_to_fspath.
- Fix setup_cursor array allocation.
- Fix free of unintialized pointer in xfs_acl_valid error path.
- Guard against path string overflows.
- Check strdup results properly in initallfs().
- Fix attribute no_change_count logic.
- Remove extraneous close() in fsrallfs().
- xfs_repair now skips the freelist scan of a corrupt agf
when in no-modify mode.
- xfs_db now skips freelist scans of corrupt agfs.
- Remove unconditional ASSERT(0) in xfs_repair.
- Reduce bb_numrecs in bno/cnt btrees when log consumes all agf space.
- Add depraction message for xfs_check.
- xfs_quota allow user or group names beginning with digits reported by
James Carter.
- Fix manpages and usage() spelling, errors and omissions.

I have placed a pre-release tarball here:


Please take a look and report any issues before next Wednesday (08 May
2013). If there are other patches which you feel are essential, now is
the time to say so.


The new lines (below) in xfs_check.sh breaks older xfstests:

   xfs_check is deprecated and scheduled for removal in June 2014.
   Please use xfs_repair -n <dev> instead.

xfstests' ./check thinks the TEST directory is inconsistent and stops.
I simply commented the lines out of the installed /usr/sbin/xfs_check
for my tests.

FSTYP         -- xfs (debug)
PLATFORM      -- Linux/x86_64 host 3.9.0-rc1+
MKFS_OPTIONS  -- -f -bsize=4096 {SCRATCH_DEV}

_check_xfs_filesystem: filesystem on {TEST_DEV} is inconsistent (c) (see
Passed all 0 tests


I'm battling this as well, coming to hazy conclusions late at night. In the _xfs_check() function in common/rc of xfstests, there was something like this:

${XFS_DB_PROG}${DBOPTS} -F -i -p xfs_check -c "check$OPTS" $1

When doing XFS tests with an external logdev (at least), things failed that way, so I threw an "echo command_about_to_run" line just above that. I got the impression that the flags "-l $TEST_LOGDEV" were being passed to xfs_db twice (not fatal), but I didn't see where the $TEST_DEV was being passed to xfs_db. It looks like xfs_db is giving back the standard usage line. Again, it was late last night, so might someone verify that the xfs_db is always called with the correct data dev or mountpoint?

Anyway, something indeed is up between the latest git xfsprogs and the latest xfstests, but my blame list hasn't been set yet. I did see the deprecation message either in the ".full" file or one of the /tmp files left behind, so you are indeed correct about that. Whether that's the only error is another matter.

BTW, it looked like xfs_check is scheduled for removal in June 2014, yet the xfstests folks are planning like it will be gone in June 2013. Is the year in the deprecation message correct?



<Prev in Thread] Current Thread [Next in Thread>