Re: Git commit amend empty emails

[Junio C Hamano <gitster@pobox.com>]

Jeff King <peff@peff.net> writes:

> However, there's something else going on. I am surprised that we allow
> empty emails at all and the code here is quite strange. The first check
> on the ident format is when we feed the data to fmt_ident to generate
> the string that goes into the commit object.  We disallow empty _names_
> there, but not empty _emails_.  I'm not sure if this is an oversight, or
> an intentional historic compatibility thing.

Looking at e27ddb6 you cited, I think we knew about historical
mistakes that allowed an empty names, but not an empty e-mail
address.  We probably have tried to kill both in one stone.

> Once upon a time, it relied only on split_ident_lane to report problems.
> But Junio's e27ddb6 (split_ident_line(): make best effort when parsing
> author/committer line, 2012-08-31) made split_ident_line more lenient,
> and introduced sane_ident_split to cover the difference. Except that it
> did more than that: besides checking whether the name is empty (which
> the original split_ident_line used to do), it also complains if the
> email is empty (which is new in that commit).

> So we now notice the empty email in this code path, but the only thing
> we do is avoid writing out the environment variables and continue. Which
> means that the actual string generated by fmt_ident (complete with empty
> email) is what goes into the commit. So why are we setting the
> environment variables at all?

I think that part was more underthinking than oversight.

We didn't want to abort the commit but we didn't want to contaminate
the environment variables with known-to-be-bad values to spread the
problem further.  But there is no guarantee that not exporting the
environment variables would give us more comformant name and e-mail
address, so that thinking is flawed.

> Here are two patches to improve this. These are on top of the
> jk/commit-date-approxidate topic, as that is where the regression was
> introduced.
>
> The first one fixes the regression and can stand by itself. The second
> fixes the GIT_AUTHOR problem, but AFAIK that has been there for years.
> So it is not as urgent, but is still maint-worthy, in my opinion.
>
>   [1/2]: commit: loosen ident checks when generating template
>   [2/2]: commit: always populate GIT_AUTHOR_* variables
>
> If we did want to truly disallow empty emails, we could do a follow-on
> 3/2 that teaches fmt_ident to reject them (that is the right place
> because it is where the validation checks for the author go, and also
> because we would probably want the same validation for the committer).
>
> But I do not think we should do that lightly. It has been this way for
> years, and clearly at least one person is depending on it. If we're
> going to change it, we might want a warning/deprecation period.
>
> -Peff