Re: GNU M4 1.4.5 test failures

[Eric Blake]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

According to Nelson H. F. Beebe on 7/18/2006 5:59 AM:
> Yesterday, I built m4-1.4.5 in 25 different build environments on
> almost that many flavors of Unix.  Almost all of them had one or more
> test failures, so I have decided not to install this release.

Thanks for the report.  I'm copying bug-m4 for the bulk of your message.

http://lists.gnu.org/archive/html/bug-m4/2006-07/msg00011.html

There is a one-line patch that fixes most of these failures (079 and 082).
 Additionally, the same bug of reading beyond the end of the string was
also present in 1.4.4, so you are no worse installing 1.4.5 than you were
with 1.4.4 (it is just that 1.4.5 added testsuite coverage of the bug, and
omits the NUL character that 1.4.4 emitted, making the bug more visible
rather than the segfault on the rare cases when reading beyond string
bounds crossed page boundaries into unmapped memory).  There are other
core dumps that 1.4.5 solves (especially if any of your machines do not
provide ecvt, in which case 1.4.4 and all earlier versions have a known
buffer overrun that allow execution of arbitrary code).

> 
> Here is what I see:
> 
>       % grep -h  -A 1 'Failed check' *.log
>       Failed checks were:
>         ./079.regexp:out ./082.patsubst:out
>       --
>       Failed checks were:
>         ./079.regexp:out
>       --
>       Failed checks were:
>         ./079.regexp:out
>       --
>       Failed checks were:
>         ./079.regexp:out
>       --
>       Failed checks were:
>         ./079.regexp:out ./082.patsubst:out
>       --
>       Failed checks were:
>         ./082.patsubst:out ./092.platform_ma:out

Which platform was this?  We were unable to determine a platform macro to
use (such as __unix__, __windows__, or __os2__).  Is there a good
preprocessor constant we can key off of to define a platform macro for
this system?

>       --
>       Failed checks were:
>         ./079.regexp:out
>       --
>       Failed checks were:
>         ./079.regexp:out ./082.patsubst:out
>       --
>       Failed checks were:
>         ./085.format:out
>       --
>       Failed checks were:
>         ./079.regexp:out ./082.patsubst:out ./092.platform_ma:out
>       --
>       Failed checks were:
>         ./082.patsubst:out
>       --
>       Failed checks were:
>         ./079.regexp:out ./082.patsubst:out ./082.patsubst:err 
> ./096.sysval:out

What platform was this?  What was the failure mode of 096.sysval:out?  Can
you attach the actual output from that command?  sysval is inherently
non-portable, since system() semantics vary on non-POSIX systems, and our
testsuite may be making a bad assumption.

>       --
>       Failed checks were:
>         ./079.regexp:out ./082.patsubst:out
>       --
>       Failed checks were:
>         ./079.regexp:out
>       --
>       Failed checks were:
>         ./079.regexp:out ./082.patsubst:out
>       --
>       Failed checks were:
>         ./079.regexp:out ./082.patsubst:out
>       --
>       Failed checks were:
>         ./082.patsubst:out
>       --
>       Failed checks were:
>         ./082.patsubst:out ./082.patsubst:err
>       --
>       Failed checks were:
>         ./079.regexp:out ./082.patsubst:out
>       --
>       Failed checks were:
>         ./079.regexp:out ./082.patsubst:out
>       --
>       Failed checks were:
>         ./079.regexp:out ./082.patsubst:out
>       --
>       Failed checks were:
>         ./079.regexp:out ./082.patsubst:out ./082.patsubst:err
> 
> I leave on a month-long vacation at this end of this week, so I likely
> won't be able to help debug this problem.
> 
> For your convenience, I've put the build logs here at this private
> unpublished URL:
> 

[stripped URL from bug-m4 post]

> Please don't distribute them or reveal the URL, since the logs contain
> machine details that could make our systems targets for directed
> attacks.
> 
> -------------------------------------------------------------------------------
> - Nelson H. F. Beebe                    Tel: +1 801 581 5254                  
> -
> - University of Utah                    FAX: +1 801 581 4148                  
> -
> - Department of Mathematics, 110 LCB    Internet e-mail: address@hidden  -
> - 155 S 1400 E RM 233                       address@hidden  address@hidden -
> - Salt Lake City, UT 84112-0090, USA    URL: http://www.math.utah.edu/~beebe/ 
> -

[Hey, I live in Salt Lake, too!]

- --
Life is short - so eat dessert first!

Eric Blake             address@hidden
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.1 (Cygwin)
Comment: Public key at home.comcast.net/~ericblake/eblake.gpg
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFEvM9W84KuGfSFAYARAp4VAJwMqNhiezWolfB+spGoUg6L+zTrswCeMqkm
CHZIg8BGckrhOW/rnynA4dc=
=iNUa
-----END PGP SIGNATURE-----