Christian Weisgerber:

> http://build-failures.rhaalovely.net/amd64/2026-06-10/

I think I'll soon mark the remaining fallout from the LLVM 22 switch
as BROKEN, so it doesn't distract from new problems in the bulk
builds.

A few random comments:

> cad/magic                   just22@atlantide.mooo.com

Upstream is very active and this should be updated.

> databases/pgtcl             stwo@users.sourceforge.net

Looks like the rest of the code freely slaps (void *) casts on
everything and they just forgot in a few places.

> games/oolite                ports@

The embedded old copy of spidermonkey does questionable things
with va_list*.  I could probably commit a workaround, but ewww.
There seem to be newer releases, but at a quick glance I couldn't
even tell if source code is available.

> japanese/Wnn                espie@
> japanese/canna              espie@

This cruft breaks on every compiler update and I'm out of patience.

> lang/v                      jasper@

Should be easy to fix for somebody who understands the V type system,
so not me, but this is also lagging behind the upstream.

> x11/rep-gtk                 ports@

I don't know how to fix this one.  The included lisp definitions
for some GTK functions disagree with the latter's actual signatures,
and I don't understand how to make them match.  Upstream Git has a
bunch of fixes, but not for the first problem the build errors out
on.  I guess we could try commenting out the offenders.  The fact
that the single dependent, the sawfish window manager, has successfully
worked despite those manifestly incorrect bindings suggests that
they aren't used.

-- 
Christian "naddy" Weisgerber                          naddy@mips.inka.de