Re: firefox115

["David H. Gutteridge" <david%gutteridge.ca@localhost>]

On Tue, 24 Jun 2025 at 22:39:59 +0200 (CEST), Havard Eidnes wrote:
> > Meanwhile my "what error do I actually get building firefox115 with
> > rust 1.86.0 on NetBSD/amd64 9.3" build is chugging away building
> > lang/clang, so it'll be a while.
> 
> Didn't get to it; the build bombed with
> 
> /usr/pkgsrc/lang/wasi-compiler-rt/work/compiler-rt-
> 19.1.7.src/lib/builtins/absvdi2.c:19:24: error: Invalid address space
> for WebAssembly target
>    19 | COMPILER_RT_ABI di_int __absvdi2(di_int a) {
>       |                        ^
> /usr/pkgsrc/lang/wasi-compiler-rt/work/compiler-rt-
> 19.1.7.src/lib/builtins/absvdi2.c:19:24: error: Invalid address space
> for WebAssembly target
> 2 errors generated.
> *** Error code 1
> 
> Stop.
> make[2]: stopped in /usr/pkgsrc/lang/wasi-compiler-rt/work/compiler-
> rt-19.1.7.src/cmake-pkgsrc-build
> *** Error code 1
> 
> i don't quite know what that's about...
> Hints & suggestions solicited.

This is probably fallout from the LLVM update to 19, which caused
breakage with clang on NetBSD 9.x. (This is perhaps related to PR
pkg/59439.)

I don't believe Rust is the cause of the firefox115 breakage, I believe
it began with the update to LLVM 19. firefox115 built for me with the
last state of LLVM 18 in pkgsrc plus rust-1.85.1nb2, on NetBSD
9.4_STABLE. (Since I was the last to make an update to that package,
that was the build environment for me when doing so.) After the LLVM
update to 19, Rust binding generation by firefox115 seemed different
than before, and it broke at that point. (That is, it was certainly
broken before the update to Rust 1.86 in pkgsrc HEAD.)

It's not realistic to expect upstreams will be carrying patches for the
"latest" whatever dependency with ESR releases. They're more likely to
have maintained a stablized build environment, the opposite of what
pkgsrc likes to do.

Dave