svn commit: r230353 - head/usr.sbin/makefs

[Andreas Tobler]

On 25.01.12 07:03, Hiroki Sato wrote:
> Andreas Tobler<andreast at FreeBSD.org>  wrote
>    in<4F1F94B3.6020803 at FreeBSD.org>:
>
> an>  On 25.01.12 01:40, Hiroki Sato wrote:
> an>  >  Andreas Tobler<andreast at FreeBSD.org>   wrote
> an>  >     in<4F1F3585.8060802 at FreeBSD.org>:
> an>  >
> an>  >  an>   Hi Eitan,
> an>  >  an>
> an>  >  an>   On 20.01.12 02:38, Eitan Adler wrote:
> an>  >  an>   >   Author: eadler
> an>  >  an>   >   Date: Fri Jan 20 01:38:21 2012
> an>  >  an>   >   New Revision: 230353
> an>  >  an>   >   URL: http://svn.freebsd.org/changeset/base/230353
> an>  >  an>   >
> an>  >  an>   >   Log:
> an>  >  an>   >      Fix warning when compiling with gcc46:
> an>  >  an>   >              error: variable 'temp' set but not used
> an>  >  an>   >
> an>  >  an>   >      Approved by:	dim
> an>  >  an>   >      Approved by: cperciva (mentor, blanket for pre-mentorship
> an>  >  an>   >      already-approved commits)
> an>  >  an>   >      MFC after:	3 days
> an>  >  an>
> an>  >  an>   I do not know which of the makefs commits it was:
> an>  >  an>
> an>  >  an>  [andreast at neon] /export/home/andreast/>  makefs -t cd9660 -o
> an>  >  chrp-boot
> an>  >  an>  -o rockridge -o label=pseries -B4321 p.iso
> an>  >  /data1/netboot/powerpc64/
> an>  >  an>   Segmentation fault (core dumped)
> an>  >  an>
> an>  >  an>   [neon:~] andreast% uname -ra
> an>  >  an>   FreeBSD neon.andreas.nets 10.0-CURRENT FreeBSD 10.0-CURRENT #11
> an>  >  an>   r230469M: Mon Jan 23 02:53:05 CET 2012
> an>  >  an>
> an>  >  andreast at neon.andreas.nets:/usr/obj/export/devel/fbsd/head/src/sys/NEON
> an>  >  an>   amd64
> an>  >  an>
> an>  >  an>   Reverting to 230352 lets me create an iso.
> an>  >  an>
> an>  >  an>   I compile makefs with the base compiler.
> an>  >  an>
> an>  >  an>   I'll do some more investigations tomorrow, late night here.
> an>  >
> an>  >    I got the same symptom and am investigating it.  Can you rebuild the
> an>  >    binary with a debug option like this:
> an>  >
> an>  >    % cd /usr/src/usr.sbin/makefs
> an>  >    % make clean
> an>  >    % make DEBUG_FLAGS=-g
> an>  >    % make DEBUG_FLAGS=-g install
> an>  >
> an>  >    and then send me the output of the following command?
> an>  >
> an>  >    % printf "run -t cd9660 -o chrp-boot -o rockridge -o label=pseries
> an>  >    % -B4321 p.iso /data1/netboot/powerpc64/\nbt\nf 1\n f 2\n" | gdb -x
> an>  >    % /dev/stdin -batch /usr/sbin/makefs
> an>  >
> an>  >    In my environment both the old and the new version could reproduce
> an>  >    it.  I am not sure if mine is the same as yours at this moment,
> an>  >    though.
> an>
> an>  It is actually r230354, this is the commit which shows the failure.
> an>  And I reverted back to 230353 and onfirmed that it 'works'.
>
>   Thank you!  I will investigating it.  Actually my test data set can
>   make SIGSEGV in both of the versions for some reason.  I have to
>   narrow down the cause, anyway.
>
>   I guess your /data1/netboot/powerpc64/ is quite large, but can I
>   receive the tarball of it in some way?  It would be helpful.

It is a powerpc64 crossbuild installation, built on amd64. It is around 
650MB unzipped, 230MB zipped. I can upload it to freefall later this 
evening.

Thanks,
Andreas