[darcs-users] Re: darcs annotate format obscures the code

John S. Yates, Jr. jyates at netezza.com
Mon Mar 21 20:53:31 UTC 2005


> If you use the patch date encoded in PT as the change key
> then they're both in date order and tolerant to inserts.

Is no one worried about skew between timestamps generated on
different machines?  In theory, really bad skew could cause
patches to be displayed in an order contradicting recorded
dependencies.

I believe that order of integration is the crucial factor in
tracking down history.  Order of authoring is a red herring.
For example, consider the following scenario:

  - Devo A generates a local patch P.
  - Devo A goes off-line (sick? vacation?) for a while.
  - Devo B goes on a coding binge and generates a large
    number of local patches.
  - Devo A comes back on-line and sends patch P upstream.
  - Devo B continues his coding binge and generates even more
    local patches.
  - Project leader L gets around to integrating P and
    along the way resolves a conflict it introduced.
  - Devo B pulls from the upstream repository.
  - Devo B's repository now exhibits a regression which he
    needs to track down.

Do you really think B wants to see patches collated by their
independent creation date?

/john







More information about the darcs-users mailing list