On Mon, 27 Apr 2020 at 17:51, Richard Purdie
<richard.purdie@linuxfoundation.org> wrote:

On Mon, 2020-04-27 at 16:13 +0300, Samuli Piippo wrote:

meta-mingw is using Mingw-w64 but the host triplet does not match
the expected pattern. config/os/mingw32-w64 enables more features
that were not working with mingw32.

Signed-off-by: Samuli Piippo <samuli.piippo@qt.io>
---
 meta/recipes-devtools/gcc/gcc-9.3.inc         |  1 +
 ...nfig-os-mingw32-w64-for-mingw-builds.patch | 36 +++++++++++++++++++
 2 files changed, 37 insertions(+)
 create mode 100644 meta/recipes-devtools/gcc/gcc-9.3/0040-Always-use-config-os-mingw32-w64-for-mingw-builds.patch

diff --git a/meta/recipes-devtools/gcc/gcc-9.3.inc b/meta/recipes-devtools/gcc/gcc-9.3.inc
index b0411078d3..4c8275bf2a 100644
--- a/meta/recipes-devtools/gcc/gcc-9.3.inc
+++ b/meta/recipes-devtools/gcc/gcc-9.3.inc
@@ -67,6 +67,7 @@ SRC_URI = "\
            file://0037-CVE-2019-14250-Check-zero-value-in-simple_object_elf.patch \
            file://0038-gentypes-genmodes-Do-not-use-__LINE__-for-maintainin.patch \
            file://0039-process_alt_operands-Don-t-match-user-defined-regs-o.patch \
+           file://0040-Always-use-config-os-mingw32-w64-for-mingw-builds.patch \
 "
 S = "${TMPDIR}/work-shared/gcc-${PV}-${PR}/gcc-${PV}"
 SRC_URI[sha256sum] = "71e197867611f6054aa1119b13a0c0abac12834765fe2d81f35ac57f84f742d1"
diff --git a/meta/recipes-devtools/gcc/gcc-9.3/0040-Always-use-config-os-mingw32-w64-for-mingw-builds.patch b/meta/recipes-devtools/gcc/gcc-9.3/0040-Always-use-config-os-mingw32-w64-for-mingw-builds.patch
new file mode 100644
index 0000000000..5a167f252c
--- /dev/null
+++ b/meta/recipes-devtools/gcc/gcc-9.3/0040-Always-use-config-os-mingw32-w64-for-mingw-builds.patch
@@ -0,0 +1,36 @@
+From 052f7e5db0d793155015530b639c4e42d37eb7eb Mon Sep 17 00:00:00 2001
+From: Samuli Piippo <samuli.piippo@qt.io>
+Date: Fri, 23 Aug 2019 10:10:43 +0300
+Subject: [PATCH] Always use config/os/mingw32-w64 for mingw builds
+
+Make sure we use the w64 version of the configs, since that's what
+meta-mingw is using.
+
+Signed-off-by: Samuli Piippo <samuli.piippo@qt.io>
+---
+ libstdc++-v3/configure.host | 12 ++----------
+ 1 file changed, 2 insertions(+), 10 deletions(-)
+
+diff --git a/libstdc++-v3/configure.host b/libstdc++-v3/configure.host
+index 155a3cdea..adbda8784 100644
+--- a/libstdc++-v3/configure.host
++++ b/libstdc++-v3/configure.host
+@@ -277,16 +277,8 @@ case "${host_os}" in
+     os_include_dir="os/hpux"
+     ;;
+   mingw32*)
+-    case "$host" in
+-      *-w64-*)
+-        os_include_dir="os/mingw32-w64"
+-        error_constants_dir="os/mingw32-w64"
+-        ;;
+-      *)
+-        os_include_dir="os/mingw32"
+-        error_constants_dir="os/mingw32"
+-        ;;
+-    esac
++    os_include_dir="os/mingw32-w64"
++    error_constants_dir="os/mingw32-w64"
+     OPT_LDFLAGS="${OPT_LDFLAGS} \$(lt_host_flags)"

It might make things a bit more obvious and upstreamable to just change
      *-w64-*)
to
      *-w64*)
if I understand correct?

With meta-mingw, the host triplet is actually either
i686-pokysdk-mingw32 or x86_64-pokysdk-mingw32, depending on bitness
in the SDKMACHINE value.

Hmm, it does indeed look like we are not correctly setting the host in a way it can be detected by GCC. It's a little surprising this hasn't caused any actual problems (at least up to this point)... is this something new to gcc 9.3, and do we have any idea why the AB MinGW tests aren't seeing any problems?

Either way, it would probably be better to figure out a way to make the host look correct in meta-mingw if at all possible.

Your commit message is also a little confusing because the first few times I read it, I thought we were configuring for 64-bit when 32-bit is desired, but based on the current GCC code being replaced and the value of the host-triplets, it would seem to be the opposite (we are getting 32-bit when 64-bit is desired).