On Fri, Feb 11, 2005 at 07:44:08AM +0000, Sisyphus wrote:
In version 5.45 MM_Win32 has the same as MM_Win95 (except no mention of $DMAKE). So, somewhere between then and now ....
What's *really* disappointing is that the same fix won't work with 'dmake' - starting to look like a 'dmake' deficiency. I tried the MM_Win95 form (minus the 'if $DMAKE' conditional) and that doesn't do anything for 'dmake' either.
That may explain the split. Since Makefile.PL can't really know what make you're trying to use I think the assumption has been made that Win9x is going to be using nmake and everything else maybe dmake maybe not. Its not a very good assumption but it does account for the anomaly.
Yes ... maybe that's how it happened - though one might as well assume that 'nmake' is being used, and have sub xs_o() return accordingly. That won't help 'dmake', but afaict it won't hurt 'dmake' either - it seems to simply have no effect at all.
One day someone might work out the appropriate 'dmake' fix - and *then* we'll need to know whether 'nmake' or 'dmake' is being used.
The best solution may well be to re-structure the PDL source (if it's possible) so that this issue does not arise - but one probably needs a basic understanding of MakeMaker and Makefiles to do that. (Counts *me* out :-)
I've written this up on rt.cpan.org - ID [cpan #11470].
Cheers, Rob
