git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: Don't Call commit-msg Hooks With Empty Commit Messages

From
Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Date
Sep 17, 2021, 19:27 UTC
Message-ID
<874kajdmsh.fsf@evledraar.gmail.com>
In-Reply-To
<xmqqlf3vilnb.fsf@gitster.g>
On Fri, Sep 17 2021, Junio C Hamano wrote:
Show 33 quoted lines
> Kurt von Laven <kurt.von.laven@gmail.com> writes:
>
>> The most common reason commit messages are left empty is to abort
>> them. commit-msg hooks that replace empty commit messages with
>> non-empty ones (i) make it impossible to abort commits, (ii) are
>> startling to developers joining a project configured in this manner,
>> and (iii) can offer no value that wouldn't be equally or better
>> offered another way. For instance, a default commit message would be
>> better implemented as a commit message template or prepare-commit-msg
>> hook. I propose that Git eventually cease calling commit-msg hooks
>> when the commit-message is empty, but I would understand if backwards
>> compatibility were the overriding concern. On the other hand, the
>> empty commit message case is easy to overlook when crafting a
>> commit-msg hook. One consequence of this behavior is that running the
>> popular pre-commit tool (https://pre-commit.com/) tends to lead to a
>> spew of false positives to the console on an aborted commit when
>> configured with commit-msg hooks.
>
> The primary reason commit-msg hook is there is *not* because we need
> a way to tweak the log message.  As you said, prepare-commit-msg and
> templates are much better way to give some sort of default.  
>
> The purpose of the hook is to serve as the gatekeeper to cause an
> attempt with a bad commit message to fail.  And a properly written
> commit-msg hook would be rejecting an empty message, instead of
> inserting cruft into an empty message file.
>
> So, from that point of view, if we were to change anything, a more
> useful thing to do might be to forbid commit-msg hook from modifying
> the file and make sure it would only verify but not modify, I
> suspect.  Doing so would have a side effect of making sure that no
> commit-msg hook will turn an empty message file into a non-empty
> message file ;-).

I'd think we'd want to call it on an empty message, e.g. maybe someone depends on that with empty message = auto-generate a message for me.

But for those that don't, doesn't the default behavior of "git commit" catch this in either case, i.e. it wouldn't let it pass without --allow-empty-message.

I understood this report as the hook taking the empty message (e.g. the user using it as a shorthand to abort), and their hook "helpfully" inserting some "default" message or template.

Previous: Junio C HamanoNext: Junio C Hamano
Message 3 of 6 in “Don't Call commit-msg Hooks With Empty Commit Messages”
  1. Kurt von LavenSep 17, 2021
  2. Junio C HamanoSep 17, 2021
  3. Ævar Arnfjörð BjarmasonSep 17, 2021
  4. Junio C HamanoSep 17, 2021
  5. Kurt von LavenSep 18, 2021
  6. Kurt von LavenSep 19, 2021

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.