mdadm 3.2.5: SEGFAULTS caused by sha1-includes.diff patch!

[Sebastian Riemer]

On 24.09.2012 16:08, Michael Tokarev wrote:
> Hmm. I think I already did the backport to squeeze, and it is even
> uploaded to the archive on Sept-17.  Is there somethig wrong with
> that backport (not counting the issue with sha1 includes) ?

There is no particular reason - we just prefer building from git sources
and have our own Debian repositories as well as build system. I couldn't
find any tags or branches related to a backport of 3.2.5 to Squeeze at
http://anonscm.debian.org/gitweb/?p=pkg-mdadm/mdadm.git;a=summary .

> I can't reproduce it.  I tried compiling it on wheezy and on squeeze, on
> 32bit and 64bit userspace - with and without the patch (with ansidecl.h
> from binutils-dev installed) the code produced is exactly the same.
>
> Please give me a way to reproduce this.

Ooops, it turned out to be one of our patches I thought that it wouldn't
be hit upon --zero-superblock.

Sorry! Thanks for testing - my fault!

We're working on MD RAID-1 based replication. This is why we need the
bitmap on iSCSI/SRP/iSER volumes. The bitmap is quite sophisticated. MD
RAID-1 replication is much much better than slow DRBD. But MD RAID-1
needs some little modifications esp. in terms of usability for this
purpose and to reload kernel caches in case of double assembly.

Cheers,
Sebastian