Bug 221381

+++ This bug was initially created as a clone of Bug #209283 +++

DIR=`mktemp -d`
cd $DIR
for f in {0,1,2,3,4}{0,1,2,3,4,5,6,7,8,9}
do
      echo $f
      touch
f${f}{0,1,2,3,4,5,6,7,8,9}{0,1,2,3,4,5,6,7,8,9}{0,1,2,3,4,5,6,7,8,9}{0,1,2,
3,4,5,6,7,8,9}
done

cd $DIR
ls *
<segfault>

-- Additional comment from jakub on 2006-10-04 06:48 EST --
This sounds familiar:

http://sources.redhat.com/bugzilla/show_bug.cgi?id=3253

2006-09-30  Ulrich Drepper  <drepper>

        * posix/glob.c (glob_in_dir): Add some comments and asserts to
        explain why there are no leaks.

2006-09-25  Jakub Jelinek  <jakub>

        [BZ #3253]
        * posix/glob.c (glob_in_dir): Don't alloca one struct globlink at a
        time, rather allocate increasingly bigger arrays of pointers, if
        possible with alloca, if too large with malloc.

http://sourceware.org/cgi-bin/cvsweb.cgi/libc/posix/glob.c.diff?cvsroot=glibc&r1=1.69&r2=1.72

-- Additional comment from jakub on 2006-10-16 03:43 EST --
Should be fixed in glibc-2.3.4-2.28 in dist-4E-qu-candidate.

-- Additional comment from bnocera on 2006-12-19 12:19 EST --
With glibc-2.3.4-2.28 and bash-3.0-19.3, on an i386, I get a quite different
crash now:
#0  __gconv_transform_utf8_internal (step=0x9df9c30, data=0xbf40011c, 
    inptrp=0xbf400110, inend=0x9dfe974 "/tmp.Bqlwb28746/", outbufstart=0x0, 
    irreversible=0xbf400114, do_flush=0, consume_incomplete=1)
    at ../iconv/skeleton.c:401
#1  0x00180beb in *__GI___mbrtowc (pwc=0xbf40018c, s=0x9dfe964 "*", n=16, 
    ps=0xbf400190) at mbrtowc.c:76
#2  0x080c2fdb in xdupmbstowcs (destp=0xbf4001c8, indicesp=0x0, 
    src=0x9dfe964 "*") at xmbsrtowcs.c:188
#3  0x080c2cc9 in xstrmatch (pattern=0x9dfe964 "*", 
    string=0x9e0c96f "f370729", flags=5) at smatch.c:371
#4  0x080bfe09 in glob_vector (pat=0x9dfe964 "*", 
    dir=0x9dfe970 "/tmp/tmp.Bqlwb28746/", flags=0) at glob.c:489
#5  0x080c060b in glob_filename (pathname=0x9dfe950 "/tmp/tmp.Bqlwb28746/*", 
    flags=0) at glob.c:782
#6  0x08086cb3 in shell_glob_filename (
    pathname=0x9e02d10 "/tmp/tmp.Bqlwb28746/*") at pathexp.c:254
#7  0x08083a89 in expand_word_list_internal (list=Variable "list" is not available.
) at subst.c:7200
#8  0x0806a8f2 in execute_command_internal (command=0x9e03090, asynchronous=0, 
    pipe_in=-1, pipe_out=-1, fds_to_close=0x9dfad38) at execute_cmd.c:2797
#9  0x0806c670 in execute_command (command=0x9e03090) at execute_cmd.c:351
#10 0x0805dd7b in reader_loop () at eval.c:146
#11 0x0805ce42 in main (argc=1, argv=0xbff954a4, env=0xbff954ac) at shell.c:714

Bash issue?

-- Additional comment from bnocera on 2007-01-03 10:38 EST --
Created an attachment (id=144715)
bash-segv.patch

Patch from Yves Begrand <yves.begrand>:

- backported changes appearing in glob_vector() routine on bash-3.0-31
- backported changes brought with patch bash31-016
- removed the use of alloca() routine and associated test, because:
 from alloca(3):  "The alloca function is machine and compiler dependent.
			On many systems its implementation is buggy.
			Its use is discouraged"

I've mentioned to Yves that it would be better to not make any changes
regarding alloca usage, as it would make the codebase drift more from upstream.

The patch fixes the problem encountered.