Re: c-stack [was: M4 1.4.14 on AIX 7.1BETA]

[Eric Blake]

On 07/29/2010 10:59 AM, Eric Blake wrote:
> On 07/29/2010 10:46 AM, Rainer Tammer wrote:
>>
>> And test-c-stack also dumps:
>>
>> # dbx test-c-stack
>> Type 'help' for help.
>> reading symbolic information ...
>> (dbx) where
>> __start() at 0x10000150
>> (dbx)
> 
> Ouch.  Did AIX stack overflow handling change between 6 and 7?  Do you
> have GNU libsigsegv installed?  It's optional for M4, but this test
> behaves differently depending on whether the library is present.
> 
> If libsigsegv is installed, then does it pass its own 'make check'?  It
> may be that no one has yet ported stack overflow handling to AIX 7.
> 
> If libsigsegv is not installed, but installing it passes 'make check',
> then it is just a matter of figuring out why libsigsegv is able to work
> to make gnulib's c-stack do likewise.

Wait a minute.  In your 'make -k check' output, I see:


Checking ./stackovf.test
Stack soft limit set to 300K
Pass
...
PASS: test-c-stack.sh
PASS: test-c-stack2.sh

I don't see any failures related to c-stack proper.  Debugging
test-c-stack triggers a SEGFAULT (and may point to a limitation in dbx,
if that doesn't handle graceful debugging through stack overflow and
signal handlers), but the mere act of a fault _doesn't_ mean the test
failed.  The test is all about whether the executable can still recover
enough to give a decent death message rather than a core dump.

Which means I'm not sure why you were complaining about test-c-stack -
is there some other failure that you saw but I'm missing?

Then again, since test-c-stack2.sh passed, rather than skipped, I'm
assuming you have might already have libsigsegv installed.  If so, it
would also be interesting to see the test results with libsegsegv not
present.

-- 
Eric Blake   address@hidden    +1-801-349-2682
Libvirt virtualization library http://libvirt.org

signature.asc
Description: OpenPGP digital signature