This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch to nuke POSIX and MAXPATHLEN from xm-*.h
On Sat, Mar 10, 2001 at 02:47:32AM +0100, Marc Espie wrote:
>
> Sigh... no, please don't remove OpenBSD files like this.
> I tried somewhat hard to make things symetric on those architectures.
> The `avoid surprises' comment is there for a reason. We got bitten
> by this once. I don't want to get bitten again. Note that, if you don't
> set xm_file, you will get the cpu default xm_file later on.
> This would be fine if the `cpu default file' DID reflect the cpu.
Thing is, the direction of all this work is toward not having xm-*.h
files at all. I'm testing a patch right now that kills another 40 of
them, leaving us with
alpha/xm-vms.h i386/xm-osf1elf.h rs6000/xm-beos.h
arm/xm-arm.h i386/xm-sco.h rs6000/xm-darwin.h
convex/xm-convex.h i386/xm-sco5.h rs6000/xm-mach.h
i370/xm-mvs.h i386/xm-vsta.h rs6000/xm-rs6000.h
i370/xm-oe.h m68hc11/xm-m68hc11.h sparc/xm-linux.h
i386/xm-beos.h m68k/xm-next.h vax/xm-vax.h
i386/xm-cygwin.h m68k/xm-plexus.h vax/xm-vms.h
i386/xm-djgpp.h m88k/xm-m88k.h xm-interix.h
i386/xm-isc.h mips/xm-iris5.h xm-linux.h
i386/xm-mingw32.h ns32k/xm-pc532-min.h
i386/xm-next.h romp/xm-romp.h
and I could squash even more, but the patch is getting unwieldy. All
of these do rather esoteric things which are unlikely to be of concern
to a sane, modern OS like OpenBSD.
> As things stand, the `cpu default file' often reflects the most common OS
> on that architecture as well (to wit: alpha).
There are only seven of these left with the patch I'm testing. Alpha
is not one of them.
arm/xm-arm.h
convex/xm-convex.h
m68hc11/xm-m68hc11.h
m88k/xm-m88k.h
romp/xm-romp.h
rs6000/xm-rs6000.h
vax/xm-vax.h
xm-m68hc11.h shouldn't exist at all - it is a cross-compiler target
only. xm-arm.h, xm-convex.h, and xm-vax.h only set the floating-point
format, which is not OS-specific (and will go away as soon as I have a
sane autoconf test for that).
xm-m88k.h, xm-romp.h, and xm-rs6000.h do indeed do OS- or compiler-
specific things. xm-m88k.h and xm-romp.h only do them inside OS- and
compiler-specific #ifdefs; still wrong, but unlikely to cause
problems.
That leaves xm-rs6000.h, which is quite AIX specific, and
unconditionally so. Unfortunately this is the one OpenBSD supports.
I should just rename the file rs6000/xm-aix.h.
> And I already mentioned you don't quite have all the files to all OpenBSD
> configurations, I'll try to send the remaining patches, but
> right now, I'm scrambling with a lot of things, including understanding
> why libstdc++-v3 won't build with anything but gnu-make.
I suggest you hold off until this series of patches settles down, then
reexamine what you have. You may find you need to do a lot less work
in your host configuration.
zw