Re: subversion ruby bindings and clang 15 (was: Re: amd64-clang bulk build report (llvm 15))
On Tue, Jan 24, 2023 at 12:15 PM Stefan Sperling wrote: > On Tue, Jan 24, 2023 at 07:22:24PM +0100, Jeremie Courreges-Anglas wrote: > > On Sun, Jan 08 2023, Jeremie Courreges-Anglas wrote: > > > > [...] > > > > >> > https://wxcvbn.org/~jca/build-failures/amd64-clang/2023-01-03/devel/subversion.txt > > > subversion/bindings/swig/ruby/svn_client.c:6473:17: error: > incompatible pointer to integer conversion passing 'apr_hash_t *' (aka > 'struct apr_hash_t *') to parameter of type 'VALUE' (aka 'unsigned long') > [-Wint-conversion] > > > if (NIL_P(arg2)) { > > > ^~~~ > > > > Lots of similar failures in here. I didn't try to fix it, but an > > alternate way to handle it would be to drop the ruby bindings for the > > time being (they're not used elsewhere in the tree). Something like the > > diff below. Thoughts? > > This error is in generated code. I guess this means any port that > depends on devel/swig to generate code is potentially failing with > clang 15? > > Anyway, fine with me to disable SVN ruby bindings by default if it helps. > OK jeremy@ for disabling the SVN ruby bindings as well. If devel/swig + ruby is an issue with clang 15, audio/ruby-taglib may also be worth testing. Thanks, Jeremy
Re: subversion ruby bindings and clang 15 (was: Re: amd64-clang bulk build report (llvm 15))
On Tue, Jan 24, 2023 at 07:22:24PM +0100, Jeremie Courreges-Anglas wrote: > On Sun, Jan 08 2023, Jeremie Courreges-Anglas wrote: > > [...] > > >> https://wxcvbn.org/~jca/build-failures/amd64-clang/2023-01-03/devel/subversion.txt > > subversion/bindings/swig/ruby/svn_client.c:6473:17: error: incompatible > > pointer to integer conversion passing 'apr_hash_t *' (aka 'struct > > apr_hash_t *') to parameter of type 'VALUE' (aka 'unsigned long') > > [-Wint-conversion] > > if (NIL_P(arg2)) { > > ^~~~ > > Lots of similar failures in here. I didn't try to fix it, but an > alternate way to handle it would be to drop the ruby bindings for the > time being (they're not used elsewhere in the tree). Something like the > diff below. Thoughts? This error is in generated code. I guess this means any port that depends on devel/swig to generate code is potentially failing with clang 15? Anyway, fine with me to disable SVN ruby bindings by default if it helps. > Index: Makefile > === > RCS file: /home/cvs/ports/devel/subversion/Makefile,v > retrieving revision 1.197 > diff -u -p -r1.197 Makefile > --- Makefile 13 Nov 2022 15:28:25 - 1.197 > +++ Makefile 15 Jan 2023 00:27:27 - > @@ -45,7 +45,7 @@ EXTRACT_SUFX= .tar.bz2 > > PSEUDO_FLAVORS= no_bindings no_ap2 no_perl no_python no_ruby > no_gnome-keyring > FLAVORS= maintainer_mode > -FLAVOR?= > +FLAVOR?= no_ruby > .if ${FLAVOR:Mno_bindings} > FLAVOR += no_perl no_python no_ruby > .endif > @@ -163,6 +163,8 @@ CONFIGURE_ARGS+=--without-swig > .if ${BUILD_PACKAGES:M-ruby} > MODULES+=lang/ruby > CONFIGURE_ENV+= RUBY=ruby${MODRUBY_BINREV} RDOC=rdoc${MODRUBY_BINREV} > +.else > +CONFIGURE_ENV+= RUBY=false RDOC=false > .endif > > .if ${BUILD_PACKAGES:M-ap2} > @@ -191,9 +193,11 @@ pre-patch: > post-patch: > @${SUBST_CMD} ${WRKSRC}/Makefile.in \ > ${WRKSRC}/subversion/libsvn_subr/auth.c > +.if ${BUILD_PACKAGES:M-ruby} > mv > ${WRKSRC}/subversion/bindings/swig/ruby/libsvn_swig_ruby{,${MODRUBY_BINREV}} > sed -i 's/libsvn_swig_ruby/libsvn_swig_ruby${MODRUBY_BINREV}/g' \ > `fgrep -lr libsvn_swig_ruby ${WRKSRC}` > +.endif > > .if ${BUILD_PACKAGES:M-python} > TEST_DEPENDS+= ${FULLPKGNAME-python}:${BUILD_PKGPATH},-python > > -- > jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE
subversion ruby bindings and clang 15 (was: Re: amd64-clang bulk build report (llvm 15))
On Sun, Jan 08 2023, Jeremie Courreges-Anglas wrote: [...] >> https://wxcvbn.org/~jca/build-failures/amd64-clang/2023-01-03/devel/subversion.txt > subversion/bindings/swig/ruby/svn_client.c:6473:17: error: incompatible > pointer to integer conversion passing 'apr_hash_t *' (aka 'struct apr_hash_t > *') to parameter of type 'VALUE' (aka 'unsigned long') [-Wint-conversion] > if (NIL_P(arg2)) { > ^~~~ Lots of similar failures in here. I didn't try to fix it, but an alternate way to handle it would be to drop the ruby bindings for the time being (they're not used elsewhere in the tree). Something like the diff below. Thoughts? Index: Makefile === RCS file: /home/cvs/ports/devel/subversion/Makefile,v retrieving revision 1.197 diff -u -p -r1.197 Makefile --- Makefile13 Nov 2022 15:28:25 - 1.197 +++ Makefile15 Jan 2023 00:27:27 - @@ -45,7 +45,7 @@ EXTRACT_SUFX= .tar.bz2 PSEUDO_FLAVORS=no_bindings no_ap2 no_perl no_python no_ruby no_gnome-keyring FLAVORS= maintainer_mode -FLAVOR?= +FLAVOR?= no_ruby .if ${FLAVOR:Mno_bindings} FLAVOR += no_perl no_python no_ruby .endif @@ -163,6 +163,8 @@ CONFIGURE_ARGS+=--without-swig .if ${BUILD_PACKAGES:M-ruby} MODULES+= lang/ruby CONFIGURE_ENV+=RUBY=ruby${MODRUBY_BINREV} RDOC=rdoc${MODRUBY_BINREV} +.else +CONFIGURE_ENV+=RUBY=false RDOC=false .endif .if ${BUILD_PACKAGES:M-ap2} @@ -191,9 +193,11 @@ pre-patch: post-patch: @${SUBST_CMD} ${WRKSRC}/Makefile.in \ ${WRKSRC}/subversion/libsvn_subr/auth.c +.if ${BUILD_PACKAGES:M-ruby} mv ${WRKSRC}/subversion/bindings/swig/ruby/libsvn_swig_ruby{,${MODRUBY_BINREV}} sed -i 's/libsvn_swig_ruby/libsvn_swig_ruby${MODRUBY_BINREV}/g' \ `fgrep -lr libsvn_swig_ruby ${WRKSRC}` +.endif .if ${BUILD_PACKAGES:M-python} TEST_DEPENDS+= ${FULLPKGNAME-python}:${BUILD_PKGPATH},-python -- jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE signature.asc Description: PGP signature