Good and bad news.

Ryan C. Gordon icculus at
Tue Jul 23 21:06:26 EDT 2002

The good: PhysicsFS is pretty darn portable. Thanks to the TestDrive
servers at Compaq, I could try it out on some odd hardware/OS combinations.

With minor tweaks, I got the library running on:

   FreeBSD: Alpha and x86
   OpenBSD: Alpha and x86
   Linux: Itanium and Alpha.

I tried NetBSD, but they don't have pthread support...I'm not sure if this
means we can do without or NetBSD purely built on fork()ing, or
does it have threads of some sort?

I also tried HPUX but had no luck whatsoever...I figure this might be
doable if there's a GNU toolchain for the OS, but the test servers didn't
have it.

Similar results from Compaq Tru64.

I tried to build on OpenVMS, but I couldn't figure what I was doing once I
logged in.  :)

PhysicsFS has, in the past, built on Solaris, and probably still will, but
I couldn't verify it...anyone have a shell account on a Solaris box I can

And, for kicks, I got physfs limping on AtheOS. (Minus CD-ROM detection,
since there's no CD-ROM support for AtheOS, and thread support, since
pthread.h wouldn't compile.)  They really need to clean up their build
tools, but for a tiny download, I was willing to throw it up in vmware and
try it.

Generally, none of these Operating Systems are interesting enough for me
to fall over myself to get the library building, but it was interesting to
give it a try.

Generally, there were three places I had trouble:
 - Building at all (lack of supported compilers, etc)
 - pthread support (missing headers/libs, or special command lines needed).
 - CD-ROM support (every OS does this differently. I hate that.)

At least we know the library is 64-bit clean, now.

And the bad:

The file enumeration code in zip.c was just broken, broken, broken. I've
commented it out in CVS, and will rewrite it to not suck soon.


More information about the physfs mailing list