D6183: copies: add config option for writing copy metadata to file and/or changset

[martinvonz (Martin von Zweigbergk)]

martinvonz added a comment.


  In https://phab.mercurial-scm.org/D6183#90470, @gracinet wrote:
  
  > Hi Martin,
  >
  > Thanks for taking on copy tracing, it's been on our mind for a while, too.
  >
  > Some of our users would be very interested in the expected speedups of the copy tracing system, however the impact of putting that data in the changeset itself would not be acceptable to them in practice. For instance, if I understand correctly, it would affect all existing hashes and could lead to subtle problems while exchanging data. It's possible that some of that could be worked around over time, but from our perspective, it looks as if a cache-based system would avoid them entirely. This would make the benefits of your work available to all users in the short term.
  >
  > We understand it looks to be more complex to you, but we're willing to help. I'm pretty confident that, working together, we can nail this before the freeze in a way that would lift all concerns. This matter is important enough to us that we're ready to make working with you on that a priority in our schedule.
  >
  > If that suits you, we could have a video chat this week to get over the details - of which we could post a summary here to keep the whole community in the loop.
  >
  > What do you think?
  
  
  I see benefits of both solutions. As you said, the cache solution's primary benefit is that it works on existing repos. For new repos (in environments where everyone using the repo has a new Mercurial version), it seems simpler to store and exchange the copy information in the changeset instead of writing it to the filelog and also to a cache.
  
  I think the way I've written these patches, it shouldn't be too hard for you guys to extend it (in core) by adding another option for `experimental.copies.read-from` to make it read from the cache. I'm happy to talk about that. I'm honestly not happy to talk about doing only the cache solution. Okay with you?

REPOSITORY
  rHG Mercurial

REVISION DETAIL
  https://phab.mercurial-scm.org/D6183

To: martinvonz, #hg-reviewers
Cc: gracinet, marmoute, mercurial-devel