Re: [PATCH] Try unbreaking devel/libexecinfo on mips64*

Index: Makefile
===================================================================
RCS file: /cvs/ports/devel/libexecinfo/Makefile,v
retrieving revision 1.26
diff -u -p -r1.26 Makefile
--- Makefile 3 Jan 2020 15:16:59 -0000 1.26
+++ Makefile 25 Oct 2020 10:24:23 -0000
@@ -9,7 +9,7 @@ GH_PROJECT= backtrace
DISTNAME = ${GH_PROJECT}-${V}
PKGNAME = libexecinfo-$V
EPOCH = 0
-REVISION = 2
+REVISION = 3
CATEGORIES = devel

SHARED_LIBS = execinfo 2.0
@@ -27,7 +27,8 @@ FAKE_FLAGS = PREFIX=${WRKINST}/${TRUEPRE
MANDIR=${PREFIX}/man/man

.if ${MACHINE_ARCH:Mmips64*} || ${MACHINE_ARCH:Mhppa} || ${MACHINE_ARCH:Msh}
-CFLAGS+= -D__BUILTIN_HACK
+COMPILER= base-gcc base-clang
+CFLAGS+= -D__BUILTIN_HACK -fno-omit-frame-pointer
.endif

do-configure:
On 10/25/2020, Brad Smith wrote:
> From ports-gcc Makefile...
>
> BUILD_DEPENDS += devel/bison \
> devel/libexecinfo
>

Ah, I gave a quick look at LIB_DEPENDS, but not at BUILD_DEPENDS,
my bad.

Then, if the __BUILTIN_HACK part is only possible with GCC at the
moment, and if we can't use ports-gcc, then I see that adding
-fno-omit-frame-pointer makes the test pass with gcc-4.2.1, too.

octeon still has base-gcc I believe, so with that ports-gcc
constraint, that's the only thing I see to let that "hack" build.

Updated diff, tested on loongson again.

(Of couse, if it's still wrong, no problem. I'm just trying
to help unbreaking that part of the mips64* ports tree.)