On Sat, May 25, 2019 at 01:12:06PM -0400, Brian Callahan wrote:
> On 5/25/19 11:58 AM, Leonid Bobrov wrote:
> > Index: Makefile.inc
> > ===================================================================
> > RCS file: Makefile.inc
> > diff -N Makefile.inc
> > --- /dev/null 1 Jan 1970 00:00:00 -0000
> > +++ Makefile.inc 25 May 2019 15:51:13 -0000
> > @@ -0,0 +1,17 @@
> > +# $OpenBSD: Makefile,v 1.17 2019/05/03 16:59:10 danj Exp $
> > +
> > +ONLY_FOR_ARCHS = i386 amd64 sparc64 mips64 mips64el powerpc
>
> How did you come up with this list? Either it should be all archs with
> backend code in pcc (in which case you're missing some archs) or it should
> be all archs that are known to build (that would be amd64 i386 mips64
> mips64el) or it should be all archs that are known to build and actually
> work (amd64 i386). You should use the build and actually work list.
>
This list comes from configure scripts (and pkgsrc don't even have this
list in lang/pcc-current, while in lang/pcc they set it to arm*, i386,
mips*, powerpc, sparc64, vax and x86_64):
pcc configure (NetBSD has longer list)
```
openbsd*)
targos=openbsd
abi=elf
stabs=yes
case "$target_cpu" in
i?86) targmach=i386 ;;
vax) targmach=vax ;;
mips64el) targmach=mips64 ;;
mips64) targmach=mips64 endian=big ;;
powerpc) targmach=powerpc endian=big ;;
sparc64) targmach=sparc64 endian=big ;;
m68k) targmach=m68k endian=big ;;
x86_64) targmach=amd64 ;;
esac
;;
```
pcc-libs configure (NetBSD has longer list)
```
openbsd*)
targos=openbsd
case "$target_cpu" in
i?86) targmach=i386 ;;
mips64el) targmach=mips ;;
mips64) targmach=mips endian=big ;;
powerpc) targmach=powerpc endian=big ;;
sparc64) targmach=sparc64 endian=big ;;
vax) targmach=vax ;;
x86_64) targmach=amd64 ;;
esac
;;
```
I didn't add m68k and vax to this list because OpenBSD doesn't support
them (however OpenBSD supports m88k, but I don't know is it compatible
with m68k or not).
I didn't actually test that because I either need exotic hardwares or
need to setup virtual machines, so I choose to trust configure script
and to trust what NetBSD/pkgsrc already do.
> > Index: pcc/Makefile
> > ===================================================================
> > RCS file: pcc/Makefile
> > diff -N pcc/Makefile
> > --- /dev/null 1 Jan 1970 00:00:00 -0000
> > +++ pcc/Makefile 25 May 2019 15:51:13 -0000
> > @@ -0,0 +1,21 @@
> > +# $OpenBSD: Makefile,v 1.17 2019/05/03 16:59:10 danj Exp $
> > +
> > +COMMENT = portable C compiler
> > +
> > +DISTNAME = pcc-${V}
>
> This would make PKGNAME=pcc-${V}, which isn't ok. Likewise, the PKGNAME for
> pcc-libs is pcc-libs-${V} which also isn't ok.
>
> ~Brian
>
That's what I was asking in the start of this topic: should I use
versioning like PKGNAME=pcc=1.1.0.${V}?
