On Mon, 2013-03-11 at 13:08 -0400, Stephen Gallagher wrote:
On Mon 11 Mar 2013 01:03:52 PM EDT, Lukas Slebodnik wrote:
> On (11/03/13 12:38), Stephen Gallagher wrote:
>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>>
>> We were not properly detecting that cmocka was unavailable. It
>> was expecting an empty value and getting "no" instead. This
>> patch corrects the expectation, so we will now skip building and
>> running cmocka tests on platforms that do not have it available.
>>
>> Also, the common_cmocka.h header was missing from the
>> distribution tarball, causing 'make distcheck' to fail.
>>
>>
>>
>>
>>
>> Note: the nss cmocka test is consistently segfaulting for me
>> during 'make distcheck' on F18 x86_64. I haven't had time to look
>> into why, yet. It works fine in normal 'make check', so I suspect
>> there's likely something going on with permissions (since 'make
>> check' sets $libdir as read-only). But that's just a guess.
>
> Another solution to proper detection of libcmocka is attached in
> my patch. I am not very familiar with autotools, so I am not sure
> which solution is more suitable.
>
Both are probably fine. I'm reattaching my patch though, as I
accidentally sent the earlier one I had made that didn't have the
Makefile.am fixes as well.
I'm personally slightly in favor of setting the explicit default and
testing for it (the way my patch does it) just because explicit values
are easier to spot later if we're trying to track down misbehavior.
ack