[GEM-dev] linking order?

[IOhannes m zmoelnig]

james tittle wrote:
> hmm,
> 
> ...one thing I've noticed with the configure/make switch is the 
> reemergence of an old weirdness, which is that [pix_film] doesn't  work
> on osx unless pix_filmDarwin.cpp is compiled/linked previous to 
> pix_film.cpp:  why is this, and is there any way around it?

it seems to be a problem with the automagic registering of classes.
"pix_film" (which is quite dumb) should _not_ autoregister itself if
there exists an implementation (like "pix_filmDarwin"), which in turn
has to autoregister itself.
why ? :
both "pix_filmDarwin" and "pix_film" bind to [pix_film] (or at least:
they should); but if pix_film autoregisters itself before pix_filDarwin,
then [pix_film] will be already taken when the latter tries to bind
itself to it.


but at least this might bring us again near the "NEW" (argh, in the
meantime i hate this naming) implementation.

> 
> ...it's particularly puzzling because the ordering doesn't seem to 
> change the behaviour of [pix_video] & pix_videoDarwin.cpp...

yes, i just looked at the code and:
- pix_video and pix_film are somewhat different
- there is an obvious bug in both of them (parts of the code are only
executed if all "linux", "apple" and "m$" are true...)
- the #define-hacks are very ugly.



> 
> any ideas?

i have fixed the most apparent things; probably it will work....


mfg.asd.r
IOhannes