Buildworld failure in libmagic on old system

Gavin Atkinson gavin.atkinson at ury.york.ac.uk
Wed Mar 14 18:50:04 UTC 2007


On Sun, 2007-03-11 at 12:24 +0900, Pyun YongHyeon wrote:
> On Sat, Mar 10, 2007 at 06:23:53PM -0500, Kris Kennaway wrote:
>  > On Sat, Mar 10, 2007 at 11:02:01PM +0000, Gavin Atkinson wrote:
>  > > 
>  > > Hi all,
>  > > 
>  > > I've been seeing a build failure on sparc64 for a while on one of my
>  > > machines, last built Thu Dec  8 13:52:22 GMT 2005 (yes, I know it's
>  > > old!), where it looks like there's some issue with which version of libc
>  > > gets used.  I've been trying to upgrade this every two months or so
>  > > since then, without success, so it's time to report it!
>  > > 
>  > > Full compile log at http://www-users.york.ac.uk/~ga9/build-20070309.txt
>  > > 
>  > > I suspect a libc issue because of the following:
>  > > 
>  > > leeloo# pwd
>  > > /usr/obj/usr/src/lib/libmagic
>  > > leeloo# ./mkmagic magic
>  > > Segmentation fault (core dumped)
>  > > leeloo# echo $LD_PRELOAD
>  > > LD_PRELOAD: Undefined variable.
>  > > leeloo# setenv LD_PRELOAD /usr/obj/usr/src/lib/libc/libc.so.7
>  > > leeloo# echo $LD_PRELOAD
>  > > /usr/obj/usr/src/lib/libc/libc.so.7
>  > > leeloo# ./mkmagic magic
>  > > leeloo#
>  > > 
>  > > Is this something which should be fixed in the source?  I'm sure setting
>  > > the OSRELDATE=0 hack to tell it the world that is installed is "really
>  > > old" may well fix it, and I'm happy to do that, I'm just wondering if
>  > > something is not being rebuilt somewhere where it should be?
>  > > 
>  > > There is one Google hit with a similar problem -
>  > > http://lists.freebsd.org/pipermail/freebsd-sparc64/2006-June/004159.html
>  > > - which is a tinderbox failure in June 2006.  I'm pretty sure I first
>  > > saw the problem before that (although I could be wrong).  I also can't
>  > > seem to see any commit that would have got the tinderbox working.
>  > 
>  > There is a malloc error in that code which causes it to segfault when
>  > MALLOC_OPTIONS=AJ.  Not doing this is the workaround, but someone
>  > should track down and fix it.
>  > 
> 
> Long time ago I've encountered the same problem on sparc64 and I posted
> possible fix to CURRENT. However I didn't receive any comments form
> obrien@ since then. See the follow thread and posted patch.
> 
> http://lists.freebsd.org/pipermail/freebsd-current/2006-August/065115.html

Thanks, your fix from this thread worked for me.  It would be nie to see
this committed, even though it does seem like I'm the only person seeing
the problem described.

I'm still slightly confused as to why my LD_PRELOAD worked around the
issue, I guess it may be something to do with the new malloc()?

Gavin


More information about the freebsd-sparc64 mailing list