On Mon, 23 Apr 2018 20:14:47 +0200, Jeremie Courreges-Anglas wrote:
> On Tue, Apr 17 2018, Pascal Stumpf <pas...@stumpf.co> wrote:
> > On Mon, 16 Apr 2018 23:14:53 +0200, Jeremie Courreges-Anglas wrote:
> >> On Mon, Apr 16 2018, <s_g...@telus.net> wrote:
> >> > I tried to continue the build of php with the ports snapshot of today
> >> > and ran into another gcc problem:
> >>
> >> [...]
> >>
> >> > /usr/ports/pobj/gcc-4.9.4/gcc-4.9.4/gcc/config/arm/arm.md:97:16:
> >> > warning: equality comparison with
> >> > extraneous parentheses [-Wparentheses-equality]
> >> > (thumb1_code)) == (
> >> > ~~~~~~~~~~~~~~~^~~~
> >> > /usr/ports/pobj/gcc-4.9.4/gcc-4.9.4/gcc/config/arm/arm.md:97:16: note:
> >> > remove extraneous parentheses
> >> > around the comparison to silence this warning
> >> > (thumb1_code)) == (
> >> > ^
> >> > /usr/ports/pobj/gcc-4.9.4/gcc-4.9.4/gcc/config/arm/arm.md:97:16: note:
> >> > use '=' to turn this equality
> >> > comparison into an assignment
> >> > (thumb1_code)) == (
> >> > ^~
> >> > =
> >> > /usr/ports/pobj/gcc-4.9.4/gcc-4.9.4/gcc/config/arm/neon.md:3486:10917:
> >> > fatal error: bracket nesting
> >> > level exceeded maximum of 256
> >> > /usr/ports/pobj/gcc-4.9.4/gcc-4.9.4/gcc/config/arm/neon.md:3486:10917:
> >> > note: use -fbracket-depth=N to
> >> > increase maximum nesting level
> >> > 33 warnings and 1 error generated.
> >> > gmake[3]: *** [Makefile:1058: insn-attrtab.o] Error 1
> >>
> >> Well this seems to be handled by the diff in my previous mail, attached
> >> again for convenience. gcc (FLAVOR='no_java no_ada') packages, install,
> >> and seems to run. dpb ongoing.
> >>
> >> Technically a clang-specific fix but it already happened with clang-5.0.
> >
> > This may be a better approach; after all, we need the flag for only one
> > file.
>
> OTOH if a single file is affected, -fbracket-depth=n shouldn't have
> side-effects on other files...
>
> > Currently testing on armv7, will have results by tomorrow.
> >
> > The real fix would be to test for the flag in configure, of course.
>
> Yeah, or maybe clang should cope better with such files. Anyway, even
> if we can probably remove gcc-4.9 from the dependency chain of a few
> more ports on arm, it would be nice to have gcc-4.9 fixed. :)
>
> If your patch is sufficient. If not, please consider using Markus'
> patch. ok jca@
Haven't managed to get my version working properly yet. So OK for
Markus' patch.
> > Index: patches/patch-gcc_Makefile_in
> > ===================================================================
> > RCS file: /cvs/ports/lang/gcc/4.9/patches/patch-gcc_Makefile_in,v
> > retrieving revision 1.1
> > diff -u -p -r1.1 patch-gcc_Makefile_in
> > --- patches/patch-gcc_Makefile_in 10 Feb 2015 22:04:54 -0000 1.1
> > +++ patches/patch-gcc_Makefile_in 17 Apr 2018 14:57:22 -0000
> > @@ -2,9 +2,23 @@ $OpenBSD: patch-gcc_Makefile_in,v 1.1 20
> >
> > Disable fixincludes madness
> >
> > ---- gcc/Makefile.in.orig Tue Feb 3 17:16:24 2015
> > -+++ gcc/Makefile.in Tue Feb 3 17:17:57 2015
> > -@@ -2730,7 +2730,7 @@ stmp-fixinc: gsyslimits.h macro_list fixinc_list \
> > +Index: gcc/Makefile.in
> > +--- gcc/Makefile.in.orig
> > ++++ gcc/Makefile.in
> > +@@ -1944,6 +1944,12 @@ DRIVER_DEFINES = \
> > +
> > + CFLAGS-gcc.o += $(DRIVER_DEFINES)
> > +
> > ++ISCLANG = $(shell $CC --version | grep -c clang)
> > ++
> > ++ifeq ($(ISCLANG),1)
> > ++CFLAGS-insn-attrtab.o += -fbracket-depth=512
> > ++endif
> > ++
> > + specs.h : s-specs ; @true
> > + s-specs : Makefile
> > + lsf="$(lang_specs_files)"; for f in $$lsf; do \
> > +@@ -2730,7 +2736,7 @@ stmp-fixinc: gsyslimits.h macro_list fixinc_list \
> > gcc_dir=`${PWD_COMMAND}` ; \
> > export TARGET_MACHINE srcdir SHELL MACRO_LIST && \
> > cd $(build_objdir)/fixincludes && \
> >
>
> --
> jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE
