On Sun, 13 Jan 2008, James Keenan via RT wrote:
> I was studying Jarkko's patch this morning in preparation for applying
> it. I had a number of concerns.
>
> 1. It turned out that at the point the patch was originally submitted,
> HEAD had moved a bit beyond the version against which Jarkko diffed. I
> had refactored some of the patched code into internal subroutines. So I
> had to adjust the patch to reflect that.
That's fine.
> 2. If I understand the thrust of the patch correctly, it says, "Only
> set these warnings if we're using GCC as our C compiler. Otherwise,
> skip it."
Yes. That's the basic idea. (Later amended to possibly try them if
you're using icc also.)
> If that's correct, then the value which we need to consult
> inside the Parrot::Configure object at the start of runstep() is not
> options->{gccversion} but data->{gccversion}.
Fine. You probably understand the innards of the Parrot::Configure object
better than anyone else.
> If I understand
> config/auto/gcc.pm correctly, if a user has GCC set up as her default C
> compiler, then data->{gccversion} will be set to the GCC version number
> in that config step -- regardless of whether a value was set for
> 'gccversion' on the command-line. So I further revised the patch to
> reflect this understanding.
I suppose so. Normally, gccversion should indeed be set to the value
corresponding to the compiler the user will actually be using to compile
parrot. I hadn't thought at all about what a command-line override for
gccversion ought to do. Offhand, I think it's reasonable to simply ignore
it.
Anyway, the config/auto/warnings.pm part appears to work out just fine.
The t/steps/auto_warnings*.t all still have the unrelated failure due
to using init::defaults and getting the wrong compiler.
--
Andy Dougherty [EMAIL PROTECTED]
