bug#61879: [core-updates] libstdc++ fails to build on powerpc64le-linux

bug#61879: [core-updates] libstdc++ fails to build on powerpc64le-linux

[Ludovic Courtès]

From <https://ci.guix.gnu.org/build/440447/details> (commit
0825162a35362640b47d3cabe278fd7ea49140bd and later):

--8<---------------cut here---------------start------------->8---
libtool: compile:  g++ -I/tmp/guix-build-libstdc++-11.3.0.drv-0/gcc-11.3.0/libstdc++-v3/../libgcc -I/tmp/guix-build-libstdc++-11.3.0.drv-0/gcc-11.3.0/build/include/ -I/tmp/guix-build-libstdc++-11.3.0.drv-0/gcc-11.3.0/build/include -I/tmp/guix-build-libstdc++-11.3.0.drv-0/gcc-11.3.0/libstdc++-v3/libsupc++ -std=gnu++17 -nostdinc++ -fno-implicit-templates -Wall -Wextra -Wwrite-strings -Wcast-qual -Wabi=2 -fdiagnostics-show-location=once -ffunction-sections -fdata-sections -frandom-seed=fs_path.lo -mno-gnu-attribute -fimplicit-templates -c ../../../libstdc++-v3/src/c++17/fs_path.cc
../../../libstdc++-v3/src/c++17/floating_from_chars.cc: In function 'std::from_chars_result std::from_chars(const char*, const char*, __ieee128&, std::chars_format)':
../../../libstdc++-v3/src/c++17/floating_from_chars.cc:499:8: error: 'string' is not a member of 'std::pmr'; did you mean 'std::string'?
  499 |   pmr::string buf(&mr);
      |        ^~~~~~
In file included from /tmp/guix-build-libstdc++-11.3.0.drv-0/gcc-11.3.0/build/include/string:39,
                 from ../../../libstdc++-v3/src/c++17/floating_from_chars.cc:34:
/tmp/guix-build-libstdc++-11.3.0.drv-0/gcc-11.3.0/build/include/bits/stringfwd.h:79:33: note: 'std::string' declared here
   79 |   typedef basic_string<char>    string;
      |                                 ^~~~~~
../../../libstdc++-v3/src/c++17/floating_from_chars.cc:504:55: error: 'buf' was not declared in this scope
  504 |       if (const char* pat = pattern(first, last, fmt, buf)) [[likely]]
      |                                                       ^~~
make[3]: *** [Makefile:762: floating_from_chars.lo] Error 1
make[3]: *** Waiting for unfinished jobs....
In file included from ../../../libstdc++-v3/src/c++17/fs_ops.cc:58:
../../../libstdc++-v3/src/c++17/../filesystem/ops-common.h: In function 'std::filesystem::file_type std::filesystem::make_file_type(const stat_type&)':
../../../libstdc++-v3/src/c++17/../filesystem/ops-common.h:269:35: warning: unused parameter 'st' [-Wunused-parameter]
  269 |   make_file_type(const stat_type& st) noexcept
      |                  ~~~~~~~~~~~~~~~~~^~
make[3]: Leaving directory '/tmp/guix-build-libstdc++-11.3.0.drv-0/gcc-11.3.0/build/src/c++17'
make[2]: *** [Makefile:765: all-recursive] Error 1
make[2]: Leaving directory '/tmp/guix-build-libstdc++-11.3.0.drv-0/gcc-11.3.0/build/src'
make[1]: *** [Makefile:568: all-recursive] Error 1
make[1]: Leaving directory '/tmp/guix-build-libstdc++-11.3.0.drv-0/gcc-11.3.0/build'
make: *** [Makefile:493: all] Error 2
error: in phase 'build': uncaught exception:
srfi-34 #<condition &invoke-error [program: "make" arguments: ("-j" "32") exit-status: 2 term-signal: #f stop-signal: #f] 7fffefc213c0> 
phase `build' failed after 6.0 seconds
command "make" "-j" "32" failed with status 2
builder for `/gnu/store/bpx8ar4qq4bkskhhj25l5phxxji0ljf8-libstdc++-11.3.0.drv' failed with exit code 1
@ build-failed /gnu/store/bpx8ar4qq4bkskhhj25l5phxxji0ljf8-libstdc++-11.3.0.drv - 1 builder for `/gnu/store/bpx8ar4qq4bkskhhj25l5phxxji0ljf8-libstdc++-11.3.0.drv' failed with exit code 1
--8<---------------cut here---------------end--------------->8---

Ludo’.

bug#61879: Patch

[Andreas Enge]

[-- Attachment #1: Type: text/plain, Size: 343 bytes --]

Hello,

attached is a new commit in old syntax, mixing both our commits.
I have confirmed that it does not change the gcc-11 build on x86_64 and i686.
But do we need "--force" for patching?

Could you maybe check again whether this builds gcc-11 on powerpc?

If yes, we can push it, and feel free to gexpify it in a separate commit.

Andreas


[-- Attachment #2: 0001-gnu-gcc-11-Fix-build-on-powerpc64le.patch --]
[-- Type: text/plain, Size: 3837 bytes --]

From 9900f9e9b86550e7d336b04ad46fba088e28cbd6 Mon Sep 17 00:00:00 2001
From: Andreas Enge <andreas@enge.fr>
Date: Thu, 13 Apr 2023 11:46:47 +0200
Subject: [PATCH] gnu: gcc-11: Fix build on powerpc64le.

* gnu/packages/patches/gcc-11-libstdc++-powerpc.patch: New file.
* gnu/local.mk (dist_patch_DATA): Register patch.
* gnu/packages/gcc.scm (make-libstdc++): Apply patch for gcc versions >= 11
and < 12 on ppc64le.

Co-authored-by: Christopher Baines <mail@cbaines.net>
---
 gnu/local.mk                                   |  1 +
 gnu/packages/gcc.scm                           | 11 ++++++++++-
 .../patches/gcc-11-libstdc++-powerpc.patch     | 18 ++++++++++++++++++
 3 files changed, 29 insertions(+), 1 deletion(-)
 create mode 100644 gnu/packages/patches/gcc-11-libstdc++-powerpc.patch

diff --git a/gnu/local.mk b/gnu/local.mk
index b07811f1cb..1255846462 100644
--- a/gnu/local.mk
+++ b/gnu/local.mk
@@ -1184,6 +1184,7 @@ dist_patch_DATA =						\
   %D%/packages/patches/gcc-11-libstdc++-hurd-libpthread.patch   \
   %D%/packages/patches/gcc-12-cross-environment-variables.patch \
   %D%/packages/patches/gcc-10-tree-sra-union-handling.patch	\
+  %D%/packages/patches/gcc-11-libstdc++-powerpc.patch           \
   %D%/packages/patches/gcolor3-update-libportal-usage.patch	\
   %D%/packages/patches/gd-fix-tests-on-i686.patch		\
   %D%/packages/patches/gd-brect-bounds.patch			\
diff --git a/gnu/packages/gcc.scm b/gnu/packages/gcc.scm
index a511cdbc45..ae324219d3 100644
--- a/gnu/packages/gcc.scm
+++ b/gnu/packages/gcc.scm
@@ -2,7 +2,7 @@
 ;;; Copyright © 2012-2023 Ludovic Courtès <ludo@gnu.org>
 ;;; Copyright © 2014, 2015, 2018 Mark H Weaver <mhw@netris.org>
 ;;; Copyright © 2014, 2015, 2016, 2017, 2019, 2021 Ricardo Wurmus <rekado@elephly.net>
-;;; Copyright © 2015 Andreas Enge <andreas@enge.fr>
+;;; Copyright © 2015, 2023 Andreas Enge <andreas@enge.fr>
 ;;; Copyright © 2015-2018, 2020-2023 Efraim Flashner <efraim@flashner.co.il>
 ;;; Copyright © 2016 Carlos Sánchez de La Lama <csanchezdll@gmail.com>
 ;;; Copyright © 2018 Tobias Geerinckx-Rice <me@tobias.gr>
@@ -889,6 +889,15 @@ (define-public (make-libstdc++ gcc)
                                           ":")
                                          "\nAM_CXXFLAGS = ")))))))
                '())
+         ,@(let ((version (package-version gcc)))
+                (if (and (target-ppc64le?)
+                    (version>=? version "11")
+                    (not (version>=? version "12")))
+               `((add-after 'unpack 'patch-powerpc
+                   (lambda _
+                    (invoke "patch" "--force" "-p1" "-i"
+                            (search-patch "gcc-11-libstdc++-powerpc.patch")))))
+               '()))
          ;; Force rs6000 (i.e., powerpc) libdir to be /lib and not /lib64.
          (add-before 'chdir 'fix-rs6000-libdir
            (lambda _
diff --git a/gnu/packages/patches/gcc-11-libstdc++-powerpc.patch b/gnu/packages/patches/gcc-11-libstdc++-powerpc.patch
new file mode 100644
index 0000000000..aff2ef16f1
--- /dev/null
+++ b/gnu/packages/patches/gcc-11-libstdc++-powerpc.patch
@@ -0,0 +1,18 @@
+diff -u -r gcc-11.3.0.alt/libstdc++-v3/src/c++17/floating_from_chars.cc gcc-11.3.0/libstdc++-v3/src/c++17/floating_from_chars.cc
+--- gcc-11.3.0.alt/libstdc++-v3/src/c++17/floating_from_chars.cc	2023-04-13 11:36:08.169841428 +0200
++++ gcc-11.3.0/libstdc++-v3/src/c++17/floating_from_chars.cc	2023-04-13 11:36:54.825827304 +0200
+@@ -495,8 +495,14 @@
+ from_chars(const char* first, const char* last, __ieee128& value,
+ 	   chars_format fmt) noexcept
+ {
++#if _GLIBCXX_USE_CXX11_ABI
+   buffer_resource mr;
+   pmr::string buf(&mr);
++#else
++  string buf;
++  if (!reserve_string(buf))
++    return make_result(first, 0, {}, ec);
++#endif
+   size_t len = 0;
+   errc ec = errc::invalid_argument;
+   __try
-- 
2.39.2

bug#61879: Patch

[Andreas Enge]

Am Thu, Apr 13, 2023 at 06:57:41PM +0200 schrieb Andreas Enge:
> attached is a new commit in old syntax, mixing both our commits.
> I have confirmed that it does not change the gcc-11 build on x86_64 and i686.
> But do we need "--force" for patching?
> Could you maybe check again whether this builds gcc-11 on powerpc?

I just pushed it, since it cannot break much... If CI shows it does not
work, we can still revert it.

Andreas

bug#61879: Patch

[Christopher Baines]

[-- Attachment #1: Type: text/plain, Size: 693 bytes --]


Andreas Enge <andreas@enge.fr> writes:

> Am Thu, Apr 13, 2023 at 06:57:41PM +0200 schrieb Andreas Enge:
>> attached is a new commit in old syntax, mixing both our commits.
>> I have confirmed that it does not change the gcc-11 build on x86_64 and i686.
>> But do we need "--force" for patching?
>> Could you maybe check again whether this builds gcc-11 on powerpc?
>
> I just pushed it, since it cannot break much... If CI shows it does not
> work, we can still revert it.

I haven't tried this yet, but I've had a quick look. I'm not sure
search-patches will work where it is, since that'll be running in the
build environment, without any access to the patches in the Guix git
repository.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 987 bytes --]

bug#61879: Patch

[Andreas Enge]

Am Fri, Apr 14, 2023 at 09:20:03AM +0100 schrieb Christopher Baines:
> I haven't tried this yet, but I've had a quick look. I'm not sure
> search-patches will work where it is, since that'll be running in the
> build environment, without any access to the patches in the Guix git
> repository.

Good point. But is this not exactly like your previous commit, which
I understood you had tested? But you are right:
   https://ci.guix.gnu.org/build/909454/log/raw
error: in phase 'patch-powerpc': uncaught exception:
unbound-variable #f "Unbound variable: ~S" (search-patch) #f 
phase `patch-powerpc' failed after 0.0 seconds
I will revert (the good news: it indeed did not break any other
architecture).

If we do not have access to the patch during build, we would need to
replace it by an invocation of substitute*, with a lot of escaping of
special characters like line ends, which would be annoying to test
(well, one could test on x86_64 on a dummy package with the same source
as gcc-11).

Andreas

bug#61879: Patch

[Christopher Baines]

[-- Attachment #1: Type: text/plain, Size: 1428 bytes --]


Andreas Enge <andreas@enge.fr> writes:

> Am Fri, Apr 14, 2023 at 09:20:03AM +0100 schrieb Christopher Baines:
>> I haven't tried this yet, but I've had a quick look. I'm not sure
>> search-patches will work where it is, since that'll be running in the
>> build environment, without any access to the patches in the Guix git
>> repository.
>
> Good point. But is this not exactly like your previous commit, which
> I understood you had tested? But you are right:
>    https://ci.guix.gnu.org/build/909454/log/raw
> error: in phase 'patch-powerpc': uncaught exception:
> unbound-variable #f "Unbound variable: ~S" (search-patch) #f 
> phase `patch-powerpc' failed after 0.0 seconds
> I will revert (the good news: it indeed did not break any other
> architecture).
>
> If we do not have access to the patch during build, we would need to
> replace it by an invocation of substitute*, with a lot of escaping of
> special characters like line ends, which would be annoying to test
> (well, one could test on x86_64 on a dummy package with the same source
> as gcc-11).

The changes I muddled together differed in that the search-patches bit
was ungexp'ed, so the patch file was handled through that mechanism. The
other important change is the actual patch itself differed as well (ec
needs moving up a bit).

I've made those changes to the commit you pushed earlier and pushed to
core-updates now.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 987 bytes --]

bug#61879: Patch

[Andreas Enge]

Am Fri, Apr 14, 2023 at 11:37:05AM +0100 schrieb Christopher Baines:
> I've made those changes to the commit you pushed earlier and pushed to
> core-updates now.

That did it, lots of green dots in the dashboard. Thanks!

Closing the bug now.

Andreas