On Fri 24 Oct 2025 at 06:53:02 -0400, Greg Troxel wrote:
> Martin Husemann <martin%duskware.de@localhost> writes:
>
> > I have:
> >
> > [~builds/current/obj/external/gpl3/gcc] martin@martins > ll ./usr.bin/include/gmp.h
> > lrwxr-xr-x 1 builds builds 72 Oct 10 21:31 ./usr.bin/include/gmp.h@ -> /home/martin/current/src/external/lgpl3/gmp/lib/libgmp/arch/x86_64/gmp.h
>
> I do too, at least now, but I used to have such links in
> usr.bin/backend. I removed them, and then things built ok.
>
> So I think I had leftover .h symlinks in the *source tree* from before,
> and cvs ignores them, but somehow they mess up the build. I don't
> understand why.
I had an unexplained symlink in my source tree too recently, and it
wasn't when using cvs, but with mercurial, and I noticed during a build
after I switched branches. (But normally I build from my read-only cvs
tree so the problem may have existed earlier)
I poked around a bit, but did not get to the bottom of it. What I did
find was this Makefile fragment that looks like it could have created
the symlinks. It ought to be outdated if the source file actually
existed in the repo, as it appeared to do.
src-public/external/gpl3/gcc/usr.bin/cc1objplus/Makefile
which contains
BUILDSYMLINKS+= ${DIST}/gcc/objc/objc-act.cc objcp-act.cc \
${DIST}/gcc/objc/objc-lang.cc objcp-lang.cc
> After removing them, building gcc/usr.bin/backend worked, and then an
> update build.sh succeeded.
Similar for me.
> Overall I think there's a spurious cvsignore for symlinks that used to
> be made in the sources, and it's really a bug to ever have made them.
I wonder if there might be a BUILDSYMLINKS associated with your symlink
too?
The full mail at
https://mail-index.netbsd.org/current-users/2025/10/10/msg046891.html
contains a bit more detail.
-Olaf.
--
___ Olaf 'Rhialto' Seibert <rhialto/at/falu.nl>
\X/ There is no AI. There is just someone else's work. --I. Rose
Attachment:
signature.asc
Description: PGP signature