On Mon, Jan 19, 2009 at 6:30 AM, Martin Evans <[email protected]>wrote:
> I didn't get the impression this person was doing a single report - he is
> 16th on the leader board of test submissions. If the C compiler is not set
> up properly (and see below because I did not generate the compiler error) he
> could generate thousands of UNKNOWN for any module that has XS - does not
> seem any point in that.
>
There's a debate at the moment on #p5p about whether it's an error to have
$Config{cc} and not have the named compiler. If so, every AS perl is
"broken".
There *is* a point -- it's how well does a distribution that requires a
compiler deal with the situation where a compiler doesn't exist. Does it
die with a reasonable error to a user? Or does it fail with something
unintelligible. It's just a prerequisite like any other.
> In this case, the "UNKNOWN" result is correct (in the new definition of
> it). The build failed and tests could not be run, thus the result of tests
> is unknown. Your error message is very descriptive -- I think it's perhaps
> the best Makefile.PL error message I've ever seen.
>
> That error did not come from me - I think it may have come from
> ExtUtils::MakeMaker but I'm not sure.
>
I meant the long text from Makefile.PL about running the VCVARS32.bat file
and so on. You're being very prescriptive about how to get the compiler set
up correctly. I think that's awesome.
-- David
<[email protected]>
