On Thu, 7 Sept 2023 at 11:45, Jakub Jelinek <jakub@redhat.com> wrote:

> On Thu, Sep 07, 2023 at 10:20:19AM +0200, Christophe Lyon via Gcc-patches
> wrote:
> > On Tue, 5 Sept 2023 at 16:38, Tobias Burnus <tobias@codesourcery.com>
> wrote:
> >
> > > That's based on the fail
> > > https://gcc.gnu.org/pipermail/gccadmin/2023q3/020349.html
> > > and on the discussion on IRC.
> > >
> >
> > Sorry I didn't notice the problem, nor the discussion on IRC, but I can
> see
> > that my commits created the problem, sorry for that.
> >
> > I'm not sure how your patch would have prevented me from doing this?
> > What happened is that I had 3 patches on top of master
> > - HEAD: the one I wanted to push
> > - HEAD-1: revert of HEAD-2
>
> git reset HEAD^
> instead of committing a revert would be better I think.
>

The patch I reverted locally was still under discussion, so I wanted to
keep it around until we made a decision.
But yeah.....


> > - HEAD-2:  libstdc-Use-GLIBCXX_CHECK_LINKER_FEATURES-for.patch
> >
> > I had actually forgotten about HEAD-1 and HEAD-2, HEAD was unrelated to
> > those, so when I pushed, I pushed 3 commits while I thought there was
> only
> > one.
> > I did run contrib/gcc-changelog/git_check_commit.py (I realize I'm not
> sure
> > whether I used git_check_commit.py or git_commit.py), but only on HEAD
> > since I had forgotten about the other two.
>
> Could you please remove your
> 2023-09-04  Christophe Lyon  <christophe.lyon@linaro.org>
>
>         PR libstdc++/111238
>         * configure: Regenerate.
>         * configure.ac: Call GLIBCXX_CHECK_LINKER_FEATURES in cross,
>         non-Canadian builds.
> libstdc++-v3/ChangeLog entry if that commit is indeed not in (or add
> a Revert: entry for it right after it if you think it needs to be in)?
> That is a part I haven't done, my/Arsen's hacks to make the version update
> get through basically ignored that revert commit.
>
> ChangeLog files can be changed by commits which only touch ChangeLog files
> and nothing else (ok, date stamp too, but please don't update that), no
> ChangeLog in the message needs to be provided for such changes.
>

Like so:
 commit d2bb261dbf282bbb258e1e5f17c1b6230327e076 (HEAD -> master)
Author: Christophe Lyon <christophe.lyon@linaro.org>
Date:   Fri Sep 8 08:13:32 2023 +0000

    Revert "libstdc++: Use GLIBCXX_CHECK_LINKER_FEATURES for cross-builds
(PR111238)"

diff --git a/libstdc++-v3/ChangeLog b/libstdc++-v3/ChangeLog
index 8f7b01e0563..0c60149d7f6 100644
--- a/libstdc++-v3/ChangeLog
+++ b/libstdc++-v3/ChangeLog
@@ -4,6 +4,16 @@
        for freestanding.
        * configure: Regenerate.

+2023-09-04  Christophe Lyon  <christophe.lyon@linaro.org>
+
+       Revert
+       2023-09-04  Christophe Lyon  <christophe.lyon@linaro.org>
+
+       PR libstdc++/111238
+       * configure: Regenerate.
+       * configure.ac: Call GLIBCXX_CHECK_LINKER_FEATURES in cross,
+       non-Canadian builds.
+
 2023-09-04  Christophe Lyon  <christophe.lyon@linaro.org>

        PR libstdc++/111238

I inserted my "Revert"  ChangeLog entry between the entry I want to declare
reverted and Jonathan's more recent patch about GLIBCXX_ENABLE_BACKTRACE.
Is that OK for the commit hooks?


        Jakub
>
>