From: Marc Espie <marc.espie.openbsd@gmail.com>
Subject: Re: regress test for the gcc15 issue
To: ports@openbsd.org
Date: Tue, 23 Jun 2026 17:18:58 +0200

On Tue, Jun 23, 2026 at 04:39:06PM +0200, Kirill A. Korinsky wrote:
> On Tue, 23 Jun 2026 16:23:44 +0200,
> Marc Espie <marc.espie.openbsd@gmail.com> wrote:
> > 
> > Here is a simple reproducer for the issue
> > 
> > Index: Makefile
> > ===================================================================
> > RCS file: /build/data/openbsd/cvs/src/regress/usr.sbin/pkg_add/Makefile,v
> > diff -u -p -r1.90 Makefile
> > --- Makefile	14 Sep 2025 18:41:11 -0000	1.90
> > +++ Makefile	23 Jun 2026 14:22:49 -0000
> > @@ -13,14 +13,14 @@ REGRESS_TARGETS=pkgnames pkgpaths signat
> >  	tying-1 tying-2 tag-delete setsprint temp-error1 \
> >  	temp-error2 temp-error2-as-root temp-error3 temp-error4 \
> >  	cksum-error system-version XXX-warning symlink no-symlink \
> > -	oldlib2 lib-error
> > +	oldlib2 lib-error gcc15
> >
> 
> And after switch to gcc16 or gcc17... shall it be renamed?
> 
> >  REGRESS_ROOT_TARGETS = temp-error2-as-root
> >  
> >  # lots of tests that really require a vstat overhaul
> >  REGRESS_EXPECTED_FAILURES = collision-check1 collision-check2 \
> >  	collision-check5 collision-check6 collision-check7 \
> > -	exotest tying-2
> > +	exotest tying-2 gcc15
> >  
> >  PERLSCRIPT = perl -I ${.CURDIR} ${.CURDIR}
> >  
> > @@ -52,7 +52,7 @@ D$i ?= ${.OBJDIR}/dest$i
> >  DEST$i ?= ${D$i}/usr/local
> >  .endfor
> >  
> > -# current maxes (used): D49, SRC30, plist28, rep7
> > +# current maxes (used): D49, SRC30, plist28, rep9
> >  pkgnames:
> >  	${PERLSCRIPT}/check-name
> >  
> > @@ -338,7 +338,8 @@ rep0/gb-0.tgz rep1/gb-0p0.tgz rep0/gd-0.
> >  rep0/gf-1.tgz rep1/gf-0.tgz \
> >  rep1/missc-0.tgz rep0/q5-0.tgz rep1/q5-1.tgz \
> >  rep0/ma-0.tgz rep1/ma-1.tgz rep0/mc-0.tgz rep1/me-1.tgz \
> > -rep0/tcl-8.5.7.tgz:
> > +rep0/tcl-8.5.7.tgz rep8/gcc-libs-15.tgz rep9/gcc-libs-15.tgz \
> > +rep9/gcc-libs-16.tgz:
> >  	@${CREATE_PKG} ${EMPTY} $@
> >  
> >  rep0/md-0.tgz:
> > @@ -783,6 +784,8 @@ rep7/avahi-glib-1.tgz: plist26
> >  rep7/cups-1.tgz:
> >  	@${CREATE_PKG} ${WAVE_PLIST} -P't/avahi-libs:avahi-libs-*:avahi-libs-0' -P't/avahi-glib:avahi-glib-*:avahi-glib-1' -W avahi.0.0 ${EMPTY} $@
> >  
> > +rep8/cfitsio-0.tgz rep9/cfitsio-0.tgz:
> > +	@${CREATE_PKG} -P't/gcc-libs-15:gcc-libs->=15,<16:gcc-libs-15' ${EMPTY} $@
> >  
> >  rep0/sym-0.tgz: plist19
> >  	@mkdir -p ${SRC30}
> > @@ -837,6 +840,13 @@ lib-error: plist28
> >  	@rm -f lib-error.out
> >  	@if ${CREATE_PKG} -B src30 -f plist28 bad-package-0 2>lib-error.out; then exit 1; fi
> >  	@diff -u lib-error.out ${.CURDIR}/lib-error.ref
> > +
> > +gcc15: rep8/cfitsio-0.tgz rep8/gcc-libs-15.tgz \
> > +	rep9/cfitsio-0.tgz rep9/gcc-libs-15.tgz rep9/gcc-libs-16.tgz
> > +	@-rm -rf ${D50}
> > +	@ROOT=${D50} ${ADD_PKG} rep8/cfitsio-0.tgz
> > +	@ROOT=${D50} PKG_PATH=${.OBJDIR}/rep9 ${ADD_PKG} -u 2>&1|head -100 |tee error-loop.out
> > +	@if [ `wc -l <error-loop.out` -eq 100 ]; then exit 1; fi
> >  
> >  clean:
> >  	-rm -rf rep* dest* plist* src* signatures.out *.ref *.out
> > 
> 
> -- 
> wbr, Kirill
> 
> 
Nope, it's a regression test. It exemplifies a specific issue, that's all