ANNOUNCE: p11-kit 0.9

[Andreas Metzler]

On 2011-12-09 Stef Walter <stefw at collabora.co.uk> wrote:
> Whoops, just saw your email now. Silly spam filter :S

> On 2011-11-26 16:21, Andreas Metzler wrote:
> > On 2011-11-14 Stef Walter <stefw at collabora.co.uk> wrote:
> >> Announcing a new release of p11-kit
> > [...]
> > 
> > Hello,
> > This release features a new test error on Linux/i386:
> > 
> > There was 1 failure:
> > 1) test__p11_hash_set_get_clear: hash-test.c:183: should be null

> It's strange I can't duplicate this on Linux i386. Are you building with
> any special options?
[...]

Hello,
No, I see this with
./configure && make && make check
on Debian/sid on all i386 archs (linux, kfreebsd and GNU/hurd).
Debian/stable does not show the issue.

After digging a little bit deeper I am stumped. Rebuilding *only*
hash-test.o with -O0 (instead of -O2) stops the error message.
Inserting
printf ("check is %s\n", check);
before CuAssert (tc, "should be null", check == NULL);
shows this behavior:

a) standard build (i.e. -O2)
(SID)ametzler at argenau:/tmp/P11-KIT/p11-kit-0.9/tests$ ./hash-test
----------------------------------------
(SID)ametzler at argenau:/tmp/P11-KIT/p11-kit-0.9/tests$ ./hash-test
check is VALUE
....F........

There was 1 failure:
1) test__p11_hash_set_get_clear: hash-test.c:184: should be null

!!!FAILURES!!!
Runs: 13 Passes: 12 Fails: 1

(SID)ametzler at argenau:/tmp/P11-KIT/p11-kit-0.9/tests$
----------------------------------------

b) hash-test.o built with -O0
----------------------------------------
(SID)ametzler at argenau:/tmp/P11-KIT/p11-kit-0.9/tests$ rm hash-test.o ;  make CFLAGS='-O0 -Wall' hash-test.o ; make
  CC     hash-test.o
  CCLD   hash-test
(SID)ametzler at argenau:/tmp/P11-KIT/p11-kit-0.9/tests$ ./hash-test
check is (null)
.............

OK (13 tests)

(SID)ametzler at argenau:/tmp/P11-KIT/p11-kit-0.9/tests$
----------------------------------------

cu andreas