Re: Git conversion: fixing email addresses from ChangeLog files

[Segher Boessenkool <segher@kernel.crashing.org>]

On Sat, Dec 28, 2019 at 04:34:20PM +0000, Richard Earnshaw (lists) wrote:
> On 28/12/2019 14:54, Segher Boessenkool wrote:
> > On Sat, Dec 28, 2019 at 01:05:13PM +0000, Joseph Myers wrote:
> >> On Sat, 28 Dec 2019, Segher Boessenkool wrote:
> >>
> >>> On Fri, Dec 27, 2019 at 07:47:02PM +0000, Richard Earnshaw (lists) wrote:
> >>>>       1 Author: Segher Boessenkool <segher@kernel,crashing.org>
> >>>> *    730 Author: Segher Boessenkool <segher@kernel.crashing.org>
> >>>>       2 Author: Segher Boesssenkool <segher@kernel.crashing.org>
> >>>
> >>> The first and third are only in changelogs.  The second even happened
> >>> only once, afaics?
> >>>
> >>> These errors only happen in the reposurgeon conversion.
> >>
> >> This is about extracting attributions from changelogs when unambiguous 
> >> there, and then correcting mistakes or otherwise making minor variants 
> >> more uniform.
> > 
> > Yes, and I'm saying you probably shouldn't do that.
> 
> Why, for heavens sake?  Even Maxim's conversion is doing this.

No, it doesn't.  If people sometimes mispel their own name in a changelog
it does not put that mispeling as Author: in the git commit.

> > Note that these errors did not exist in the changelog in the commit
> > message, for example.
> 
> Yes, they did.  Or at least, they did at the time of the original commit.

No, they never did.  I always cut off the date/name/email line from the
changelog in the commit message.

> > Since people very often typo their own name (as the evidence shows), the
> > heuristic for deriving it should be robust against that.
> 
> And the statistics show that it's not hard to identify the odd cases and
> fix them up.  Only committers with just a single commits are really hard
> to spot since we don't have data to compare against other entries.

Sure, so do that?  :-)


Segher