compile error with STABLE-20001012

[Tom Kacvinsky]

Hi Russ,

> 
> > which gives us what we want (almost).  I know the -fno-strict-aliasing
> > has to go, but what about the defines?
> 
> The -fno-strict-aliasing can stay, actually; it won't make any difference
> one way or the other.
> 

What if perl is compiled with gcc and INN is compiled with Sun's Workshop Pro C
compiler?  Sun's compiler only emits a warning when -fno-strict-aliasing is
specified as an option, but can we guarantee that any "weird" compiler flags
from one compiler won't choke another compiler?

> The defines... hm.  They're fine if INN is compiled with large files, but
> otherwise they could cause problems.  I'd better strip them.
> 
> Could you tell me what these three commands say:
> 
>   perl -MExtUtils::EmbedE -e perl_inc

 -I/usr/local/lib/perl5/sun4-solaris-64int

>   perl -MExtUtils::Embed -e ccflags

 -fno-strict-aliasing -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64

>   perl -MExtUtils::Embed -e ccdlflags

-R /usr/local/lib/perl5/sun4-solaris-64int/CORE

> 
> If we're lucky, we can just leave one of them out; if we're not, then I'll
> need to just strip out any -D options.
> 

Considering the possible compiler mixup scenario of above, wouldn't be easier to
just extract the -I options rather than trying to strip all possible weird
compiler flags?

Regards,

Tom