Git snafu

Git snafu

[Po Lu]

When trying to commit from a brand new machine installed at my
organization, I made a mistake.  I configured the git username and
e-mail as root, and not as my user.

So now there is a commit from the default Git username on emacs.git.  Is
there a way to rectify this mistake?

I'm terribly sorry for this trouble.

Re: Git snafu

[Eli Zaretskii]

> From: Po Lu <luangruo@yahoo.com>
> Date: Tue, 25 Apr 2023 20:38:51 +0800
> 
> When trying to commit from a brand new machine installed at my
> organization, I made a mistake.  I configured the git username and
> e-mail as root, and not as my user.
> 
> So now there is a commit from the default Git username on emacs.git.  Is
> there a way to rectify this mistake?
> 
> I'm terribly sorry for this trouble.

If it's on a feature branch, you could fix the problem locally, and
then publish a new branch without the mistake, I guess.

Re: Git snafu

[Po Lu]

Eli Zaretskii <eliz@gnu.org> writes:

> If it's on a feature branch, you could fix the problem locally, and
> then publish a new branch without the mistake, I guess.

This was on master, to fix a problem I found trying to install Emacs on
that machine...

Re: Git snafu

[Robert Pluim]

>>>>> On Tue, 25 Apr 2023 16:04:26 +0300, Eli Zaretskii <eliz@gnu.org> said:

    >> From: Po Lu <luangruo@yahoo.com>
    >> Date: Tue, 25 Apr 2023 20:38:51 +0800
    >> 
    >> When trying to commit from a brand new machine installed at my
    >> organization, I made a mistake.  I configured the git username and
    >> e-mail as root, and not as my user.
    >> 
    >> So now there is a commit from the default Git username on emacs.git.  Is
    >> there a way to rectify this mistake?
    >> 
    >> I'm terribly sorry for this trouble.

    Eli> If it's on a feature branch, you could fix the problem locally, and
    Eli> then publish a new branch without the mistake, I guess.

    branch: master
    commit b8f94eb0f148e774981ce711a12db604ebd174bc
    Author: Your Name <you@example.com>
    Commit: Your Name <you@example.com>

I guess we could add a .mailmap entry 

Robert
-- 

Re: Git snafu

[Eli Zaretskii]

> From: Robert Pluim <rpluim@gmail.com>
> Cc: Po Lu <luangruo@yahoo.com>,  emacs-devel@gnu.org
> Date: Tue, 25 Apr 2023 15:13:52 +0200
> 
>     branch: master
>     commit b8f94eb0f148e774981ce711a12db604ebd174bc
>     Author: Your Name <you@example.com>
>     Commit: Your Name <you@example.com>
> 
> I guess we could add a .mailmap entry 

Or we could not care.

Re: Git snafu

[Dr. Arne Babenhauserheide]

[-- Attachment #1: Type: text/plain, Size: 536 bytes --]


Po Lu <luangruo@yahoo.com> writes:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>> If it's on a feature branch, you could fix the problem locally, and
>> then publish a new branch without the mistake, I guess.
>
> This was on master, to fix a problem I found trying to install Emacs on
> that machine...

Is there any damage done?

It’s not like we’re shipping commits as product. The history is a tool,
not the goal.

Best wishes,
Arne
-- 
Unpolitisch sein
heißt politisch sein,
ohne es zu merken.
draketo.de

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 1125 bytes --]

Re: Git snafu

[Eli Zaretskii]

> From: "Dr. Arne Babenhauserheide" <arne_bab@web.de>
> Cc: Eli Zaretskii <eliz@gnu.org>, emacs-devel@gnu.org
> Date: Tue, 25 Apr 2023 16:33:48 +0200
> 
> Is there any damage done?

Only a light damage, but easily fixable.  There, done.

> It’s not like we’re shipping commits as product. The history is a tool,
> not the goal.

Our history is shipped as ChangeLog files, so those had better be in
good order.

Re: Git snafu

[Jim Porter]

On 4/25/2023 5:38 AM, Po Lu wrote:
> When trying to commit from a brand new machine installed at my
> organization, I made a mistake.  I configured the git username and
> e-mail as root, and not as my user.
> 
> So now there is a commit from the default Git username on emacs.git.  Is
> there a way to rectify this mistake?

It won't rectify it, but we could prevent this in the future with a Git 
hook (and one that's significantly simpler than the one I added for 
change log entries).

I see that the commit author is "Your Name <you@example.com>". That's 
easy enough to block; are there any other author addresses that we 
should forbid? I seem to recall that Git sometimes authors commits as 
"local-user@local-hostname", which isn't a useful email address for the 
global internet either.

Re: Git snafu

[Arsen Arsenović]

[-- Attachment #1: Type: text/plain, Size: 1206 bytes --]


Jim Porter <jporterbugs@gmail.com> writes:

> I see that the commit author is "Your Name <you@example.com>". That's easy
> enough to block; are there any other author addresses that we should forbid? I
> seem to recall that Git sometimes authors commits as
> "local-user@local-hostname", which isn't a useful email address for the global
> internet either.

It might be viable to compare the committer against the ACLs for
emacs.git or such.  The authors can vary widely, but the committer
should usually be somebody that can write to the repository.

  ~/gnu/emacs/emacs$ git show --format=fuller b8f94e | head -n 5
  commit b8f94eb0f148e774981ce711a12db604ebd174bc
  Author:     Your Name <you@example.com>
  AuthorDate: Tue Apr 25 20:28:38 2023 +0800
  Commit:     Your Name <you@example.com>            <<<<< this one
  CommitDate: Tue Apr 25 20:28:38 2023 +0800

... or, as an example of a contribution by a non-committer:

  commit 110769d039f1e5d12b948fade23563b3e540cff3
  Author:     Sam James <sam@gentoo.org>
  AuthorDate: Mon Nov 21 22:01:37 2022 +0000
  Commit:     Eli Zaretskii <eliz@gnu.org>
  CommitDate: Tue Nov 22 14:45:09 2022 +0200

-- 
Arsen Arsenović

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 381 bytes --]

Re: Git snafu

[Po Lu]

Jim Porter <jporterbugs@gmail.com> writes:

> I see that the commit author is "Your Name <you@example.com>". That's
> easy enough to block; are there any other author addresses that we
> should forbid? I seem to recall that Git sometimes authors commits as
> "local-user@local-hostname", which isn't a useful email address for
> the global internet either.

Perhaps also U. Ser <user@example.com>.

Re: Git snafu

[Eli Zaretskii]

> From: Arsen Arsenović <arsen@aarsen.me>
> Cc: Po Lu <luangruo@yahoo.com>, emacs-devel@gnu.org
> Date: Tue, 25 Apr 2023 22:41:27 +0200
> 
> It might be viable to compare the committer against the ACLs for
> emacs.git or such.  The authors can vary widely, but the committer
> should usually be somebody that can write to the repository.

But the committer's name, as recorded in the Git logs, is really of no
special interest to us.  Only the author's name and email address are
of interest.  The access rights to the Git repository are checked by
Savannah, and if the commit was successfully pushed, that check
succeeded, regardless of what's in the log.

Re: Git snafu

[Ulrich Mueller]

>>>>> On Wed, 26 Apr 2023, Eli Zaretskii wrote:

>> It might be viable to compare the committer against the ACLs for
>> emacs.git or such.  The authors can vary widely, but the committer
>> should usually be somebody that can write to the repository.

> But the committer's name, as recorded in the Git logs, is really of no
> special interest to us.  Only the author's name and email address are
> of interest.  The access rights to the Git repository are checked by
> Savannah, and if the commit was successfully pushed, that check
> succeeded, regardless of what's in the log.

But you'd catch errors due to bad configuration (as the one in question)
where both author and committer tend to be wrong.

Re: Git snafu

[Eli Zaretskii]

> From: Ulrich Mueller <ulm@gentoo.org>
> Cc: Arsen Arsenović <arsen@aarsen.me>,
>   jporterbugs@gmail.com,
>   luangruo@yahoo.com,  emacs-devel@gnu.org
> Date: Wed, 26 Apr 2023 08:10:43 +0200
> 
> >>>>> On Wed, 26 Apr 2023, Eli Zaretskii wrote:
> 
> >> It might be viable to compare the committer against the ACLs for
> >> emacs.git or such.  The authors can vary widely, but the committer
> >> should usually be somebody that can write to the repository.
> 
> > But the committer's name, as recorded in the Git logs, is really of no
> > special interest to us.  Only the author's name and email address are
> > of interest.  The access rights to the Git repository are checked by
> > Savannah, and if the commit was successfully pushed, that check
> > succeeded, regardless of what's in the log.
> 
> But you'd catch errors due to bad configuration (as the one in question)
> where both author and committer tend to be wrong.

How would we do that?  The ACL to emacs.git is not available locally,
and neither should it be, IMO.

Re: Git snafu

[Björn Bidar]

Eli Zaretskii <eliz@gnu.org> writes:

>> But you'd catch errors due to bad configuration (as the one in question)
>> where both author and committer tend to be wrong.
>
> How would we do that?  The ACL to emacs.git is not available locally,
> and neither should it be, IMO.

You can block the commit in question with a post-receive hook I think.

Re: Git snafu

[Eli Zaretskii]

> From: Björn Bidar <bjorn.bidar@thaodan.de>
> Cc: Ulrich Mueller <ulm@gentoo.org>,  arsen@aarsen.me,
>   jporterbugs@gmail.com,  luangruo@yahoo.com,  emacs-devel@gnu.org
> Date: Wed, 26 Apr 2023 12:06:49 +0300
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> >> But you'd catch errors due to bad configuration (as the one in question)
> >> where both author and committer tend to be wrong.
> >
> > How would we do that?  The ACL to emacs.git is not available locally,
> > and neither should it be, IMO.
> 
> You can block the commit in question with a post-receive hook I think.

On Savannah?  Doesn't sound like a serious enough issue to ask
Savannah folks to install this.

Re: Git snafu

[Arsen Arsenović]

[-- Attachment #1: Type: text/plain, Size: 259 bytes --]


Eli Zaretskii <eliz@gnu.org> writes:

> How would we do that?  The ACL to emacs.git is not available locally,
> and neither should it be, IMO.

Ah, I thought these checks run on Savannah.  Never mind then.

Have a lovely day!
-- 
Arsen Arsenović

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 381 bytes --]

Re: Git snafu

[Jim Porter]

[-- Attachment #1: Type: text/plain, Size: 702 bytes --]

On 4/25/2023 1:37 PM, Jim Porter wrote:
> On 4/25/2023 5:38 AM, Po Lu wrote:
>> When trying to commit from a brand new machine installed at my
>> organization, I made a mistake.  I configured the git username and
>> e-mail as root, and not as my user.
>>
>> So now there is a commit from the default Git username on emacs.git.  Is
>> there a way to rectify this mistake?
> 
> It won't rectify it, but we could prevent this in the future with a Git 
> hook (and one that's significantly simpler than the one I added for 
> change log entries).

In fact, we can do something even simpler than that: just set the config 
option "user.useConfigOnly" to true. See attached. Does this look ok to 
everyone?

[-- Attachment #2: 0001-autogen.sh-Don-t-guess-the-user-s-name-email-when-co.patch --]
[-- Type: text/plain, Size: 620 bytes --]

From 4578355ca8ef1b375eb4eaae125d6fb8c8f52cd7 Mon Sep 17 00:00:00 2001
From: Jim Porter <jporterbugs@gmail.com>
Date: Sun, 30 Apr 2023 14:09:54 -0700
Subject: [PATCH] * autogen.sh: Don't guess the user's name/email when
 committing

---
 autogen.sh | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/autogen.sh b/autogen.sh
index 6127e7b24f4..e63de1edd19 100755
--- a/autogen.sh
+++ b/autogen.sh
@@ -313,6 +313,9 @@ hooks=
 
 git_config transfer.fsckObjects true
 
+# Don't try to guess a user name/email for commits.
+
+git_config user.useConfigOnly true
 
 # Configure 'git diff' hunk header format.
 
-- 
2.25.1

Re: Git snafu

[Eli Zaretskii]

> Date: Sun, 30 Apr 2023 14:13:50 -0700
> From: Jim Porter <jporterbugs@gmail.com>
> Cc: eliz@gnu.org
> 
> In fact, we can do something even simpler than that: just set the config 
> option "user.useConfigOnly" to true. See attached. Does this look ok to 
> everyone?

How do we know no one is using several email accounts, and would like
to avoid the annoyance of being prompted by Git?

In general, I don't think we should introduce non-trivial restrictions
just because someone makes a rare mistake.  It isn't worth the
potential fallout and annoyance of people who know what they are
doing, IMO.

Thanks.

Re: Git snafu

[Jim Porter]

On 5/1/2023 5:44 AM, Eli Zaretskii wrote:
>> Date: Sun, 30 Apr 2023 14:13:50 -0700
>> From: Jim Porter <jporterbugs@gmail.com>
>> Cc: eliz@gnu.org
>>
>> In fact, we can do something even simpler than that: just set the config
>> option "user.useConfigOnly" to true. See attached. Does this look ok to
>> everyone?
> 
> How do we know no one is using several email accounts, and would like
> to avoid the annoyance of being prompted by Git?

I *do* use several email accounts, and this setting works properly for 
me. My normal configuration is to have something like this in my 
~/.gitconfig (repeated for a few other directories):

   # Set user info for repos inside ~/src
   [includeIf "gitdir:~/src/"]
     path = ~/src/.gitconfig

Then, in my Emacs clone, I override my email address again (to this 
one). With this combination of settings, everything works exactly as 
expected with "user.useConfigOnly" set.

I can also set the author of a single commit via "git commit --author" 
with "user.useConfigOnly" set (so that option is arguably a bit of a 
misnomer; from reading the documentation[1], it really only means "don't 
guess the user information").

Similarly, if I wanted to set the *committer*, I could use:

   git -c user.name=Foo -c user.email=foo@example.com commit ...

The only time a *new* error occurs (that I can see, anyway) is if you 
completely skip configuring your user info *anywhere*. Then, "git 
commit" or "git commit --author" will fail (the latter fails because Git 
still had to guess the *committer*). In this case, you can just use "git 
-c user.name ..." as described above, and all is well.

All that said, I agree that this problem is a rare enough circumstance 
that we probably don't really *need* to make my suggested change. I 
think it's a low-risk change, but I won't be upset if we just leave 
things as they are.

[1] 
https://git-scm.com/docs/git-config#Documentation/git-config.txt-useruseConfigOnly