Re: [Valgrind-developers] 2009-03-10 03:20:04 GMT nightly build (vauxhall, x86_64, Fedora 10)

[Nicholas N. <n.n...@gm...>]

On Tue, Mar 10, 2009 at 2:48 PM, Tom Hughes <th...@cy...> wrote:
> =================================================
> ./valgrind/exp-ptrcheck/tests/pth_create.stderr.diff
> =================================================
> --- pth_create.stderr.exp       2009-03-10 03:34:11.000000000 +0000
> +++ pth_create.stderr.out       2009-03-10 03:43:22.000000000 +0000
> @@ -1,6 +1,6 @@
>
>  Invalid write of size 4
> -   at 0x........: pthread_key_create (in /...libpthread...)
> +   at 0x........: pthread_key_create (pthread_key_create.c:44)
>    by 0x........: main (pth_create.c:17)
>  Address 0x........ is 1 bytes inside the accessing pointer's
>  legitimate range, a block of size 4 alloc'd
> @@ -8,7 +8,7 @@
>    by 0x........: main (pth_create.c:14)

exp-ptrcheck/tests/filter_stderr is designed to allow for pthread
debug info to be present.  There's this filter:

# Tidy up in cases where glibc (+ libdl + libpthread + ld) have
# been built with debugging information, hence source locs are present
sed \
-e "s/ vfprintf (.*)/ .../" \
-e "s/ vsprintf (.*)/ .../" \
-e "s/ sprintf (.*)/ .../" \
-e "s/ printf (.*)/ .../" \
-e "s/ strdup (.*)/ .../" \
-e "s/ pthread_key_create.c:[0-9]*/ in \/...libpthread.../" \
-e "s/ genops.c:[0-9]*/ in \/...libc.../" |

It seems that the space at the start of the last two patterns
shouldn't be there.  A similar thing (probably in tests/filter_stderr)
would cover the exit.c/_exit.c filenames that caused most of the other
failures.

N