Re: gcc12 not building on macOS 15 aarch64 (gfortran woes)

[Chuck Cranor <chuck%ece.cmu.edu@localhost>]

On Sun, Feb 23, 2025 at 10:14:59AM -0500, Greg Troxel wrote:
> What I did was cvs up -A in mk/ and lang/gcc* in a 2024Q4 tree.  I am
> now updating whole the tree to current (as that machine is now up to
> date with Q4, and I wanted to rebuild packages for that first).  I'll
> try again.


ah, ok.  that gcc13 patch was recently updated so maybe it didn't make Q4.
i've been using pkgsrc current (not Q4).   note the patch date:


sh-3.2# fgrep '$NetBSD' patches/patch-libgcc_config.host 
$NetBSD: patch-libgcc_config.host,v 1.2 2024/12/15 03:07:57 brook Exp $
sh-3.2# 


I think there is a pkgsrc policy issue here with the way the PATCHFILES
mechanism is being used?  gcc13's Makefile.common has:

.if ${MACHINE_PLATFORM:MDarwin-*-aarch64}
BUILDLINK_TRANSFORM+=   rm:-nodefaultrpaths
CONFIGURE_ARGS+=        --disable-darwin-at-rpath
PATCHFILES=             gcc-13.3.0.diff
PATCH_SITES=            https://raw.githubusercontent.com/Homebrew/formula-patches/bda0fad/gcc/
PATCH_DIST_STRIP=       -p1
.endif

so pkgsrc will only attempt to fetch and apply gcc-13.3.0.diff
on Darwin aarch64.

that leaves us in the state where we've got two different diffs
trying to patch the same file (in conflicting ways), but only on
the Darwin aarch64 platform.   Non Darwin aarch64 platforms will
patch cleanly.

if you are not a Darwin aarch64 pkgsrc developer and put a new patch
in patches/patch-*, how are you supposed to detect that it conflicts 
with a patch in PATCHFILES (i.e. gcc-13.3.0.diff)?



>   - Given the above, we should maybe/probably flip the gfortran default
>     on macos aarch64 to 14, but I think it would be much better to fix
>     gcc12.


either way is ok with me.   i don't know how to fix gcc12.


chuck