Hi Akim.
On 05/29/2012 07:23 PM, Akim Demaille wrote:
>
> Le 23 mai 2012 à 11:34, Stefano Lattarini a écrit :
>
>> + This is not a big deal in practice, since as of today, all the real
>> + world packages out there actually use 'test-driver' for most (if not
>> + all) of their tests (the introduction of TAP support in mainstream
>> + automake being too recent for those packages to have had time to phase
>> + out their protocol-less tests, assuming they wanted to -- which is
>> + usually not the case either).
>
> Actually I don't even know what benefit there is from using
> this "standard"
>
To name just a few of them:
- one test script can have several results (very useful to increase the
granularity of tests, as Automake does with, e.g., 't/instspc.tap');
- it allows to declare *conditional* expected failure (e.g., "this test
failed, but we are on AIX, so that is not a regression");
- there are sophisticated, independent runners for TAP-based tests (the
'prove' utility being one of the most known and feature-rich);
- there are feature-rich, mature perl modules that can generate TAP
output ('Test::Simple' and 'Test::More' at least, plus a plethora
of others);
- TAP is used in the testsuite of most perl modules, of Perl itself,
and also by Git (in the last couple of years at least); so it is
here to stay (de-facto standard).
> (even though I google'd, and browsed the Automake documentation (btw, there
> is no index for TAP)).
>
Regards,
Stefano
