On Mon, Mar 03, 2003 at 10:38:21PM -0500, Benjamin A. Okopnik wrote:
> On Mon, Mar 03, 2003 at 07:45:03PM -0500, Branden Robinson wrote:
[...]
> > FYI, I cannot reproduce this problem on PowerPC:
[...]
> I'm not in the least surprised; on a different architecture, the call
> stack/data stack are going to be in different places, so my location
> would be in a different place from yours. FTM, it may not even happen at
> all on yours, but you might want to try this:
>
> perl -we'$a = "a" x 100000; `MANPATH=$a xman`'
>
> If it segfaults, it's an even broader bug report. :)
No, it just silently exited with status 0.
> Just as an additional hint, here's the last line of "strace -f"
> output:
>
> --------------------------------------------------------------
> 20632 open("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
>
> <about 8000 "a"s elided>
>
> aaaaaaaa", O_RDONLY) = -1 ENAMETOOLONG (File name too long)
> 20632 --- SIGSEGV (Segmentation fault) @ 0 (0) ---
> 20632 +++ killed by SIGSEGV +++
> --------------------------------------------------------------
>
> Looks like it tries to open the contents of MANPATH without doing a
> bounds check... tisk, tisk. *Bad* "xman"... or maybe "libc". :)
Okay. Not saying it isn't a bug; just getting a handle on the scope.
--
G. Branden Robinson | Men use thought only to justify
Debian GNU/Linux | their wrong doings, and speech only
branden@debian.org | to conceal their thoughts.
http://people.debian.org/~branden/ | -- Voltaire
Attachment:
pgptChw7qmhjz.pgp
Description: PGP signature