Re: mariadb on aarch64 after the llvm 13 update (Re: aarch64 bulk build report)
On 2021 Dec 27 (Mon) at 23:32:31 +0100 (+0100), Jeremie Courreges-Anglas wrote:
:On Mon, Dec 20 2021, phess...@openbsd.org wrote:
:> bulk build on arm64.ports.openbsd.org
:> started on Mon Dec 20 03:38:26 MST 2021
:> finished at Mon Dec 20 22:29:47 MST 2021
:> lasted 0D18h51m
:> done with kern.version=OpenBSD 7.0-current (GENERIC.MP) #1455: Sun Dec 19
14:58:17 MST 2021
:>
:> built packages:8238
:> Dec 20:8237
:>
:>
:> critical path missing pkgs:
http://build-failures.rhaalovely.net/aarch64/2021-12-20/summary.log
:>
:> build failures: 10
:> http://build-failures.rhaalovely.net/aarch64/2021-12-20/databases/mariadb.log
:
:[...]
:
:mariadb tries to use "out-of-line atomics" as implemented here:
:
: https://reviews.llvm.org/D91157
:
:The summary mentions implementations in libgcc and libcompiler-rt, and
:a quick diff between compiler-rt-11.1.0 and compiler-rt-13.0.0 indeed
:lists changes that seem relevant.
:
:To work around this problem until a solution is found in base, the
:untested patch below may help. If testing goes well, ok?
:
this builds and packages on arm64, OK
:
:Index: patches/patch-configure_cmake
:===
:RCS file: patches/patch-configure_cmake
:diff -N patches/patch-configure_cmake
:--- /dev/null 1 Jan 1970 00:00:00 -
:+++ patches/patch-configure_cmake 27 Dec 2021 22:29:08 -
:@@ -0,0 +1,17 @@
:+$OpenBSD$
:+
:+libcompiler-rt from base doesn't provide support for out-of-line atomics
:+llvm commit: https://reviews.llvm.org/D91157
:+
:+Index: configure.cmake
:+--- configure.cmake.orig
: configure.cmake
:+@@ -70,7 +70,7 @@ IF(CMAKE_SYSTEM_PROCESSOR STREQUAL "i686" AND CMAKE_CO
:+ ENDIF()
:+
:+ # use runtime atomic-support detection in aarch64
:+-IF(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64")
:++IF(FALSE)
:+ MY_CHECK_AND_SET_COMPILER_FLAG("-moutline-atomics")
:+ ENDIF()
:+
:
:
:--
:jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE
--
It is impossible to make anything foolproof because fools are so
ingenious.
Re: Adding BASE_C99 for ports needing -std=gnu99 for base-gcc
On Mon, Dec 27, 2021 at 09:50:04PM -0500, Kurt Mosiejczuk wrote:
> I find myself adding the following to many ports files nowadays
>
> .include
> .if !${PROPERTIES:Mclang}
> CFLAGS += -std=gnu99
> .endif
>
> Rather than continue this, I'd like to propose adding BASE_C99 to bsd.port.mk
>
> When BASE_C99= Yes, if on a non-base-clang arch, it will add the same to
> CFLAGS, but without including bsd.port.arch.mk by hand.
>
> Included are patches to add this to bsd.port.mk and a diff for bsd.port.mk(5)
> to include it in the man page.
>
> Once this is in, I'll sweep the tree for the boilerplate and replace it all
> with this.
>
> ok?
Is it time to consider changing the gcc default to gnu99?
built and installed on sparc64 but otherwise untested.
--- /tmp/defs.old Sun Dec 19 15:25:52 2021
+++ /tmp/defs.new Sun Dec 19 16:34:59 2021
@@ -70,6 +70,7 @@
#define __FLT_HAS_INFINITY__ 1
#define __DEC64_MAX__ 9.999E384DD
#define __DEC64_MANT_DIG__ 16
+#define __STDC_VERSION__ 199901L
#define __DEC32_MAX_EXP__ 96
#define __DEC128_DEN__ 0.1E-6143DL
#define __LDBL_MANT_DIG__ 113
Index: gnu/gcc/gcc/c-opts.c
===
RCS file: /cvs/src/gnu/gcc/gcc/c-opts.c,v
retrieving revision 1.3
diff -u -p -r1.3 c-opts.c
--- gnu/gcc/gcc/c-opts.c15 Sep 2011 12:19:12 - 1.3
+++ gnu/gcc/gcc/c-opts.c28 Dec 2021 03:44:20 -
@@ -236,6 +236,8 @@ c_common_init_options (unsigned int argc
if (c_language == clk_c)
{
+ set_std_c99 (false /* ISO */);
+
/* If preprocessing assembly language, accept any of the C-family
front end options since the driver may pass them through. */
for (i = 1; i < argc; i++)
Adding BASE_C99 for ports needing -std=gnu99 for base-gcc
I find myself adding the following to many ports files nowadays
.include
.if !${PROPERTIES:Mclang}
CFLAGS += -std=gnu99
.endif
Rather than continue this, I'd like to propose adding BASE_C99 to bsd.port.mk
When BASE_C99= Yes, if on a non-base-clang arch, it will add the same to
CFLAGS, but without including bsd.port.arch.mk by hand.
Included are patches to add this to bsd.port.mk and a diff for bsd.port.mk(5)
to include it in the man page.
Once this is in, I'll sweep the tree for the boilerplate and replace it all
with this.
ok?
--Kurt
Index: bsd.port.mk
===
RCS file: /cvs/ports/infrastructure/mk/bsd.port.mk,v
retrieving revision 1.1564
diff -u -p -r1.1564 bsd.port.mk
--- bsd.port.mk 14 Dec 2021 14:21:41 - 1.1564
+++ bsd.port.mk 23 Dec 2021 16:34:42 -
@@ -736,6 +736,13 @@ CFLAGS += ${CDIAGFLAGS}
CXXFLAGS += ${CXXDIAGFLAGS}
.endif
+# Add -std=gnu99 to CFLAGS if requested and arch is base-gcc
+BASE_C99 ?= no
+.if !${PROPERTIES:Mclang} && ${BASE_C99:L} == "yes"
+CFLAGS += -std=gnu99
+.endif
+
+
# XXX trick ports into trying to write into / instead of the current user's
# homedir.
PORTHOME ?= /${PKGNAME}_writes_to_HOME
Index: bsd.port.mk.5
===
RCS file: /cvs/src/share/man/man5/bsd.port.mk.5,v
retrieving revision 1.550
diff -u -p -r1.550 bsd.port.mk.5
--- bsd.port.mk.5 22 Nov 2021 11:17:39 - 1.550
+++ bsd.port.mk.5 28 Dec 2021 02:34:51 -
@@ -966,6 +966,21 @@ Can be set to
.Sq autogen.sh
if such a script is available.
Defaults to autoreconf --force --install.
+.It Ev BASE_C99
+Defaults to
+.Sq \ .
+Set to
+.Sq Yes
+for ports needing c99 constructs while using base-gcc.
+.Pp
+When
+.Ev BASE_C99
+is set to
+.Sq Yes ,
+.Ev CFLAGS
+has
+.Sq -std=gnu99
+added to it if building on a base-gcc architecture.
.It Ev BASE_PKGPATH
Full
.Xr pkgpath 7
[sparc64] Fix build of x11/i3status
x11/i3status uses C11 constructs so base-gcc arches must use ports-gcc.
This diff fixes the build for sparc64
ok?
(cc MAINTINER)
--Kurt
Index: Makefile
===
RCS file: /cvs/ports/x11/i3status/Makefile,v
retrieving revision 1.63
diff -u -p -r1.63 Makefile
--- Makefile6 Dec 2021 09:51:28 - 1.63
+++ Makefile27 Dec 2021 20:31:30 -
@@ -17,6 +17,10 @@ PERMIT_PACKAGE = Yes
MASTER_SITES= ${HOMEPAGE}
EXTRACT_SUFX= .tar.xz
+# C11
+COMPILER= base-clang ports-gcc
+COMPILER_LANGS=c
+
WANTLIB += c confuse yajl sndio
MODULES= devel/meson
mariadb on aarch64 after the llvm 13 update (Re: aarch64 bulk build report)
On Mon, Dec 20 2021, phess...@openbsd.org wrote:
> bulk build on arm64.ports.openbsd.org
> started on Mon Dec 20 03:38:26 MST 2021
> finished at Mon Dec 20 22:29:47 MST 2021
> lasted 0D18h51m
> done with kern.version=OpenBSD 7.0-current (GENERIC.MP) #1455: Sun Dec 19
> 14:58:17 MST 2021
>
> built packages:8238
> Dec 20:8237
>
>
> critical path missing pkgs:
> http://build-failures.rhaalovely.net/aarch64/2021-12-20/summary.log
>
> build failures: 10
> http://build-failures.rhaalovely.net/aarch64/2021-12-20/databases/mariadb.log
[...]
mariadb tries to use "out-of-line atomics" as implemented here:
https://reviews.llvm.org/D91157
The summary mentions implementations in libgcc and libcompiler-rt, and
a quick diff between compiler-rt-11.1.0 and compiler-rt-13.0.0 indeed
lists changes that seem relevant.
To work around this problem until a solution is found in base, the
untested patch below may help. If testing goes well, ok?
Index: patches/patch-configure_cmake
===
RCS file: patches/patch-configure_cmake
diff -N patches/patch-configure_cmake
--- /dev/null 1 Jan 1970 00:00:00 -
+++ patches/patch-configure_cmake 27 Dec 2021 22:29:08 -
@@ -0,0 +1,17 @@
+$OpenBSD$
+
+libcompiler-rt from base doesn't provide support for out-of-line atomics
+llvm commit: https://reviews.llvm.org/D91157
+
+Index: configure.cmake
+--- configure.cmake.orig
configure.cmake
+@@ -70,7 +70,7 @@ IF(CMAKE_SYSTEM_PROCESSOR STREQUAL "i686" AND CMAKE_CO
+ ENDIF()
+
+ # use runtime atomic-support detection in aarch64
+-IF(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64")
++IF(FALSE)
+ MY_CHECK_AND_SET_COMPILER_FLAG("-moutline-atomics")
+ ENDIF()
+
--
jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE
Re: [NEW] sysutils/promscale
Denis Fondras writes: > Promscale is an open source observability backend for metrics and > traces powered by SQL. > > Built on the robust and high-peformance foundation of PostgreSQL > and TimescaleDB, it has native support for Prometheus metrics and > OpenTelemetry traces as well as many other formats like StatsD, > Jaeger and Zipkin through the OpenTelemetry Collector and is 100% > PromQL compliant. > > For Prometheus users, Promscale provides a robust and highly scalable > long-term storage system that is 100% PromQL compliant. > > https://www.timescale.com/promscale > > > [2. application/x-tar-gz; promscale-0.7.1.tgz]... The CID stuff is yicky :D - If you are ok using that approach, more power to ya - I usually go for what is done in net/miniflux, that way you can do "make modgo-gen-modulles" and just pull the date from the new MODGO_VERSION var. Other than regexie madness it looks good to me! OK abieber@
CVS: cvs.openbsd.org: ports
CVSROOT:/cvs Module name:ports Changes by: j...@cvs.openbsd.org2021/12/27 12:22:48 Modified files: security/gnutls: Makefile Added files: security/gnutls/patches: patch-lib_accelerated_aarch64_Makefile_in Log message: Unbreak on aarch64: don't use -Wa,-march=all -march=all isn't supported by clang, and it now crashes with llvm 13. Fix snatched from the FreeBSD bugtracker. Tested by phessler@, ok phessler@ ajacoutot@ (maintainer)
CVS: cvs.openbsd.org: ports
CVSROOT:/cvs Module name:ports Changes by: j...@cvs.openbsd.org2021/12/27 12:20:53 Modified files: www/surf : Makefile Log message: add missed wantlibs
CVS: cvs.openbsd.org: ports
CVSROOT:/cvs Module name:ports Changes by: j...@cvs.openbsd.org2021/12/27 11:56:32 Modified files: www/surf : Makefile distinfo www/surf/patches: patch-Makefile patch-config_def_h patch-config_mk www/surf/pkg : PLIST Added files: www/surf/patches: patch-surf_c Log message: update to surf-2.1 with help from crystal kolipe
[sparc64] Fix build of devel/glpk
devel/glpk fails out complaining that TLS (thread local storage) is
unavailable. Switching to ports-gcc on base-gcc sparc64 fixes the build.
ok?
--Kurt
Index: Makefile
===
RCS file: /cvs/ports/devel/glpk/Makefile,v
retrieving revision 1.14
diff -u -p -r1.14 Makefile
--- Makefile8 Dec 2021 10:41:18 - 1.14
+++ Makefile27 Dec 2021 16:09:07 -
@@ -12,6 +12,9 @@ MASTER_SITES= ${MASTER_SITE_GNU:=glpk/}
# GPLv3
PERMIT_PACKAGE=Yes
+COMPILER= base-clang ports-gcc
+COMPILER_LANGS=c
+
WANTLIB= c m gmp
LIB_DEPENDS= devel/gmp
Xorg modules installation
Hi, I am working on xrdpxorg (http://xrdp.org) which builds a few Xorg modules. `make plist` does not seem to able to process them: Can't put into any plist (no applicable prefix): /usr/X11R6/lib/modules/libxorgxrdp.so ... Is there a proper way to install them? (so that pkg_add puts them into the right location) Thank you! Best regards, Ingo
Fix gnutls on aarch64 after the llvm 13 update (was: Re: aarch64 bulk build report)
On Mon, Dec 20 2021, phess...@openbsd.org wrote:
> bulk build on arm64.ports.openbsd.org
> started on Mon Dec 20 03:38:26 MST 2021
> finished at Mon Dec 20 22:29:47 MST 2021
> lasted 0D18h51m
> done with kern.version=OpenBSD 7.0-current (GENERIC.MP) #1455: Sun Dec 19
> 14:58:17 MST 2021
>
> built packages:8238
> Dec 20:8237
>
>
> critical path missing pkgs:
> http://build-failures.rhaalovely.net/aarch64/2021-12-20/summary.log
[...]
> http://build-failures.rhaalovely.net/aarch64/2021-12-20/security/gnutls.log
This crash has already been hit by the FreeBSD folks;
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=260078
-march=all seems unsupported by clang, this diff fixes the crash but
doesn't add support for -march=all
https://reviews.llvm.org/D114677
The untested diff below removes the use of -march=all to fix the build
on aarch64. Tests and ok welcome.
Index: Makefile
===
RCS file: /cvs/ports/security/gnutls/Makefile,v
retrieving revision 1.169
diff -u -p -r1.169 Makefile
--- Makefile29 May 2021 22:43:09 - 1.169
+++ Makefile27 Dec 2021 16:03:05 -
@@ -3,6 +3,7 @@
COMMENT= GNU Transport Layer Security library
V= 3.7.2
+REVISION= 0
DISTNAME= gnutls-${V}
EXTRACT_SUFX= .tar.xz
Index: patches/patch-lib_accelerated_aarch64_Makefile_in
===
RCS file: patches/patch-lib_accelerated_aarch64_Makefile_in
diff -N patches/patch-lib_accelerated_aarch64_Makefile_in
--- /dev/null 1 Jan 1970 00:00:00 -
+++ patches/patch-lib_accelerated_aarch64_Makefile_in 27 Dec 2021 16:03:05
-
@@ -0,0 +1,18 @@
+$OpenBSD$
+
+Avoid -march=all, unsupported by clang.
+Clang commit that fixes the crash but doesn't add support for -march=all:
+https://reviews.llvm.org/D114677
+
+Index: lib/accelerated/aarch64/Makefile.in
+--- lib/accelerated/aarch64/Makefile.in.orig
lib/accelerated/aarch64/Makefile.in
+@@ -2104,7 +2104,7 @@ AM_CPPFLAGS = -I$(srcdir)/../../../gl -I$(builddir)/..
+ -I$(srcdir)/../../ -I$(srcdir)/../ $(am__append_1)
+
+ #ensure that we have all aarch64 instruction sets enabled for the assembler
+-AM_CCASFLAGS = -Wa,-march=all
++#AM_CCASFLAGS = -Wa,-march=all
+ EXTRA_DIST = README
+ noinst_LTLIBRARIES = libaarch64.la
+ libaarch64_la_SOURCES = aarch64-common.c aarch64-common.h \
--
jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE
Re: audio/ncspot: update to 0.9.3
On Mon, Dec 27, 2021 at 04:39:39PM +0100, Henrik Friedrichsen wrote: > On Mon, Dec 27, 2021 at 06:25:32AM +, Klemens Nanni wrote: > > Works for me on amd64, this most probably won't fix the sparc64 build > > failure since the broken crate has not been updated. > > > > OK? > > Looks good to me, thanks for the diff. Out of curiosity, which crate is > broken on sparc64? This one error: failed to run custom build command for `ring v0.16.20` as per Date: Sat, 25 Dec 2021 04:37:16 -0700 (MST) From: k...@openbsd.org To: ports@openbsd.org Subject: sparc64 bulk build report ... http://build-failures.rhaalovely.net/sparc64/2021-12-22/audio/ncspot.log ...
Re: audio/ncspot: update to 0.9.3
On Mon, Dec 27, 2021 at 06:25:32AM +, Klemens Nanni wrote: > Works for me on amd64, this most probably won't fix the sparc64 build > failure since the broken crate has not been updated. > > OK? Looks good to me, thanks for the diff. Out of curiosity, which crate is broken on sparc64?
CVS: cvs.openbsd.org: ports
CVSROOT:/cvs Module name:ports Changes by: pas...@cvs.openbsd.org 2021/12/27 06:06:30 Modified files: shells/ksh93 : Makefile distinfo shells/ksh93/pkg: PLIST Removed files: shells/ksh93/patches: patch-bin_package patch-src_cmd_ksh93_Mamfile patch-src_cmd_ksh93_edit_history_c patch-src_lib_libast_hash_hashalloc_c patch-src_lib_libast_string_tokscan_c Log message: Update to ksh93 1.0.0-beta2. Development is now taking place at https://github.com/ksh93/ksh/, with this version already being far ahead of the nine-year-old AT release we currently have.
CVS: cvs.openbsd.org: ports
CVSROOT:/cvs Module name:ports Changes by: de...@cvs.openbsd.org 2021/12/27 04:47:21 Modified files: sysutils/kopia : Makefile distinfo modules.inc Log message: update to 0.9.8 with help from Stefan Hagen, thanks.
