----- Original Message ----- 
From: "Judd Taylor" <[EMAIL PROTECTED]>
To: "Sisyphus" <[EMAIL PROTECTED]>
Cc: "Chris Marshall" <[EMAIL PROTECTED]>; <[email protected]>;
<[email protected]>
Sent: Tuesday, August 08, 2006 7:33 AM
Subject: Re: [Pdl-porters] Announcing PDL release candidate 1: please test!


> Nutz. That's 3 for 3 giving trouble!
>
> This GD interface is pretty version specific, as it parses an include
> file to build XS access to the functions (or something like that), so if
> the version windows is linking to is different, or there's an:
>
> #ifdef MSVC
> do something else
> #else
> do something normal
> #endif
>
> then segfaults may creep into it.

'_MSC_VER' is the actual term - I know such #ifdef's do exist in the PDL
source - but only in pdlcore.c (which is generated from pdlcore.c.PL). I
can't see that being the problem.  There's no mention of '_MSC_VER' anywhere
in my GD library source, but it's to be found in a few places in
perl/lib/CORE (handy.h, win32.h, win32iop.h and win32thread.h). I'll take
another look when I get the time (and inclination).

Both perl builds (MinGW and VC7) are on the same machine, accessing the very
same GD library. Both builds produce exactly the same IO/GD.xs.

I also have a VC6 built (ActiveState) perl on the same box, and a
non-optimising version of the VC6 compiler (which I rarely use, because it
*is* non-optimising) ... I've just tested and found (to my surprise) that
IO/GD builds fine with that VC6 compiler on VC6-built perl. I don't know
whether this difference is explained in terms of "VC7 versus VC6", or in
terms of "optimising versus non-optimising". (Sometimes it would be really
handy to have a proper version  of VC6.) Actually, they're different
versions of perl, too,(5.8.4 and 5.8.8) though I regard it as unlikely that
has anything to do with it.

>
> This module hasn't been heavily tested, either so there may just be
> bugs...
>
> Did you override the build, or am I really seeing that it tries to build
> the module even if can't find the library?
>

That was using the default perldl.conf (that came with the release
candidate) - I had not edited that file at all. It might not be doing things
exactly as planned, but at least the correct conclusion is being reached
.... except for IO/GD, which ought to be found. Since IO/GD apparently
builds with VC6 and definitely builds with MinGW, we probably should fix
that auto-detection issue ..... I'll take a look at that again today if I
get time .... otherwise it will have to wait until next week.

Cheers,
Rob


_______________________________________________
Perldl mailing list
[email protected]
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl

Reply via email to