[python-committers] Python workflow quirks with mercurial and hg-git extension

[Steve Dower]

On 25Jan2018 0547, Ethan Furman wrote:
> On 01/20/2018 11:19 AM, Jesus Cea wrote:
>> I plan to come back to python development (about time!) but I truly
>> hates git. I am experimenting with mercurial + hg-git extension and it
>> is quite usable (after the initial painfully slow clone time), but I am
>> having small quirks that I would like to iron out with a fellow more
>> experienced or also interested in this approach.
>>
>> Anybody out there?.
> 
> Not experienced, but interested!

Experienced enough to have given up on using hg-git for CPython and now
I just suffer git. (Those who have ever discussed workflow with me know
that's a really big deal!)

The main thing that annoyed me was the different EOL handling: Mercurial
applies the correction on update/checkout, while git does it on commit
(or perhaps the other way around, either way, it broke many PRs).
Perhaps this affects me more than most because I'm using Windows, but
commits through hg-git would leave CRLF endings in files that git would
have already translated to LF, and so the CRLF endings would show up as
changes in the PR.

The excessive clone time was also painful. I suspect dulwich could do
with some optimisation.

Perhaps these are issues that could be fixed in hg-git? If so, I'd
happily switch back to it. But at least with the backport bot doing a
great job (thanks Mariatta!) I don't have to deal with too much
switching in git.

I suspect your quirks are different from these, but feel free to give me
the details and I'll let you know if I saw them and/or worked around
them. I still use hg-git for many of my other repos.

Cheers,
Steve