Re: unbreak boost for powerpc, powerpc64
I suggest using ${PATCHORIG} rather than hardcoding to any particular value..
--
Sent from a phone, apologies for poor formatting.
On 24 December 2023 00:53:05 George Koehler wrote:
On Sat, 23 Dec 2023 23:46:33 +0100
Theo Buehler wrote:
Please go ahead, I can cope.
I committed it. Now if anyone runs update-plist, my commit adds
cpu_relax.hpp.orig.port to this warning,
Warning: entries NOT added to boost-1.84.0p0v0:
include/boost/atomic/detail/futex.hpp.orig.port (.orig.port suffix ?)
include/boost/fiber/detail/cpu_relax.hpp.orig.port (.orig.port suffix ?)
include/boost/fiber/detail/futex.hpp.orig.port (.orig.port suffix ?)
share/b2/tools/gcc.jam.orig.port (.orig.port suffix ?)
do-install runs pax -s ':^.*\.orig$$::' to skip .orig files, but
we changed PATCHORIG to .orig.port some time ago, so
-s ':^.*\.orig\.port$$::' might be better.
Re: unbreak boost for powerpc, powerpc64
On 2023-12-23 7:52 p.m., George Koehler wrote: On Sat, 23 Dec 2023 23:46:33 +0100 Theo Buehler wrote: Please go ahead, I can cope. I committed it. Now if anyone runs update-plist, my commit adds cpu_relax.hpp.orig.port to this warning, Warning: entries NOT added to boost-1.84.0p0v0: include/boost/atomic/detail/futex.hpp.orig.port (.orig.port suffix ?) include/boost/fiber/detail/cpu_relax.hpp.orig.port (.orig.port suffix ?) include/boost/fiber/detail/futex.hpp.orig.port (.orig.port suffix ?) share/b2/tools/gcc.jam.orig.port (.orig.port suffix ?) do-install runs pax -s ':^.*\.orig$$::' to skip .orig files, but we changed PATCHORIG to .orig.port some time ago, so -s ':^.*\.orig\.port$$::' might be better. Yes. That should be updated to do as intended. Thanks.OK with me.
Re: unbreak boost for powerpc, powerpc64
On Sat, 23 Dec 2023 23:46:33 +0100 Theo Buehler wrote: > Please go ahead, I can cope. I committed it. Now if anyone runs update-plist, my commit adds cpu_relax.hpp.orig.port to this warning, Warning: entries NOT added to boost-1.84.0p0v0: include/boost/atomic/detail/futex.hpp.orig.port (.orig.port suffix ?) include/boost/fiber/detail/cpu_relax.hpp.orig.port (.orig.port suffix ?) include/boost/fiber/detail/futex.hpp.orig.port (.orig.port suffix ?) share/b2/tools/gcc.jam.orig.port (.orig.port suffix ?) do-install runs pax -s ':^.*\.orig$$::' to skip .orig files, but we changed PATCHORIG to .orig.port some time ago, so -s ':^.*\.orig\.port$$::' might be better.
Re: unbreak boost for powerpc, powerpc64
On 2023-12-23 5:44 p.m., George Koehler wrote: devel/boost is broken on powerpc and powerpc64, because picks the wrong asm syntax. This patch fixes it, ok? The REVISION bump conflicts with tb's "boost futex diff"; I will do REVISION= 1 if tb commits before me. --gkoehler https://github.com/boostorg/fiber/commit/2bd8eb02f81ab8297aa89b8cb29b0f3867b9fafe OK.
Re: unbreak boost for powerpc, powerpc64
On Sat, Dec 23, 2023 at 05:44:43PM -0500, George Koehler wrote:
> devel/boost is broken on powerpc and powerpc64, because
> picks the wrong asm syntax.
> This patch fixes it, ok?
ok
> The REVISION bump conflicts with tb's "boost futex diff";
> I will do REVISION= 1 if tb commits before me.
> --gkoehler
Please go ahead, I can cope.
>
> Index: Makefile
> ===
> RCS file: /cvs/ports/devel/boost/Makefile,v
> retrieving revision 1.139
> diff -u -p -r1.139 Makefile
> --- Makefile 19 Dec 2023 06:14:25 - 1.139
> +++ Makefile 23 Dec 2023 18:33:03 -
> @@ -11,6 +11,7 @@ COMMENT-md= machine-dependent libraries
> VERSION= 1.84.0
> DISTNAME=boost_${VERSION:S/./_/g}
> PKGNAME= boost-${VERSION}
> +REVISION=0
> EPOCH = 0
> CATEGORIES= devel
> SITES=
> https://boostorg.jfrog.io/artifactory/main/release/${VERSION}/source/
> Index: patches/patch-boost_fiber_detail_cpu_relax_hpp
> ===
> RCS file: patches/patch-boost_fiber_detail_cpu_relax_hpp
> diff -N patches/patch-boost_fiber_detail_cpu_relax_hpp
> --- /dev/null 1 Jan 1970 00:00:00 -
> +++ patches/patch-boost_fiber_detail_cpu_relax_hpp23 Dec 2023 18:33:03
> -
> @@ -0,0 +1,15 @@
> +Fix powerpc and powerpc64, where clang defines __POWERPC__ but rejects
> +Darwin's "r27" syntax with "error: invalid operand for instruction".
> +
> +Index: boost/fiber/detail/cpu_relax.hpp
> +--- boost/fiber/detail/cpu_relax.hpp.orig
> boost/fiber/detail/cpu_relax.hpp
> +@@ -59,7 +59,7 @@ namespace detail {
> + // processors
> + // extended mnemonics (available with POWER7)
> + // yield == or 27, 27, 27
> +-# if defined(__POWERPC__) // Darwin PPC
> ++# if defined(__APPLE__) // Darwin PPC
> + # define cpu_relax() asm volatile ("or r27,r27,r27" ::: "memory");
> + # else
> + # define cpu_relax() asm volatile ("or 27,27,27" ::: "memory");
unbreak boost for powerpc, powerpc64
devel/boost is broken on powerpc and powerpc64, because
picks the wrong asm syntax.
This patch fixes it, ok?
The REVISION bump conflicts with tb's "boost futex diff";
I will do REVISION= 1 if tb commits before me.
--gkoehler
Index: Makefile
===
RCS file: /cvs/ports/devel/boost/Makefile,v
retrieving revision 1.139
diff -u -p -r1.139 Makefile
--- Makefile19 Dec 2023 06:14:25 - 1.139
+++ Makefile23 Dec 2023 18:33:03 -
@@ -11,6 +11,7 @@ COMMENT-md= machine-dependent libraries
VERSION= 1.84.0
DISTNAME= boost_${VERSION:S/./_/g}
PKGNAME= boost-${VERSION}
+REVISION= 0
EPOCH =0
CATEGORIES=devel
SITES=
https://boostorg.jfrog.io/artifactory/main/release/${VERSION}/source/
Index: patches/patch-boost_fiber_detail_cpu_relax_hpp
===
RCS file: patches/patch-boost_fiber_detail_cpu_relax_hpp
diff -N patches/patch-boost_fiber_detail_cpu_relax_hpp
--- /dev/null 1 Jan 1970 00:00:00 -
+++ patches/patch-boost_fiber_detail_cpu_relax_hpp 23 Dec 2023 18:33:03
-
@@ -0,0 +1,15 @@
+Fix powerpc and powerpc64, where clang defines __POWERPC__ but rejects
+Darwin's "r27" syntax with "error: invalid operand for instruction".
+
+Index: boost/fiber/detail/cpu_relax.hpp
+--- boost/fiber/detail/cpu_relax.hpp.orig
boost/fiber/detail/cpu_relax.hpp
+@@ -59,7 +59,7 @@ namespace detail {
+ // processors
+ // extended mnemonics (available with POWER7)
+ // yield == or 27, 27, 27
+-# if defined(__POWERPC__) // Darwin PPC
++# if defined(__APPLE__) // Darwin PPC
+ # define cpu_relax() asm volatile ("or r27,r27,r27" ::: "memory");
+ # else
+ # define cpu_relax() asm volatile ("or 27,27,27" ::: "memory");
