Re: [fink-core] Package test suites

[Peter O'G. <pe...@po...>]

On Aug 31, 2006, at 1:32 AM, Matthew Sachs wrote:

> On Aug 28, 2006, at 18:28, Jean-Fran=E7ois Mertens wrote:
>
>> Looking just at a couple of pkgs (in my exp dir), I see that running
>> tests involves for a number of them, in addition to the usual make
>> [-i] [-
>> k] check or the like, also additional %c _ as, just grepping from the
>> ConfigureParams in that dir :
>
> (Moving to fink-core.)
>
> So, it sounds like this feature needs to be able to do, at a minimum,
> the following in an info file:
> *Add extra dependencies for running the tests
> *Behave differently during compilation (ConfigureParams, =20
> CompileScript)
> *Specify a script which is run immediately after building and which
> will invoke the test suite and do some degree of interpretation of
> the results (exit status 0/1 for pass/fail)
>
> Here's my idea for how to add this to the .info syntax in a backwards-
> compatible matter.  Define a new field, InfoTest, which will be
> ignored unless running in test mode (either enabled as part of -m or
> with a new option).

Your proposal sounds good to me.

> Note that the packaging reference defines "InfoN" where N is an
> integer.  InfoTest  appears to work with 0.24 CVS, but if we wanted
> to strictly adhere to the spec, this would need to be surrounded by
> Info3.  If we're going to do that, it might make sense to *not* have
> a special InfoTest block (and to define a couple of extra new fields,
> like TestSource[N]), unless we want the override mechanism.

I doubt that Info3 is necessary. We came up with InfoN because a =20
change was made to info files which caused older fink's indexers to =20
be unable to parse them, thus leaving users unable to update. This =20
albeit ugly method was considered to be the best way out of that =20
situation. As far as I'm aware, simply adding some new fields to the =20
info files should not break older fink's parsing of them, so no need =20
to go Info3.

Of course Dan will probably now write to tell me I'm wrong :)

Peter=