Today I discovered two problems with the mercurial target on my mac.
The first was that deletions (and renames) do not seem to be
occurring as specified in the changeset being replayed. From what I
can tell, repo.remove(files) doesn't have an effect when the file
exists. There's an ``unlink'' argument that seems to solve my
problem, at least. I thought I'd bring it up before just sending in
a patch in case this is intentional or if there's a better way to
deal with it.
I'll go ahead and send the patch since it does solve the problem for
me. Feel free to throw it away if there's a better fix for my problem.
The second is more of a filesystem-specific problem. My filesystems
are currently case-insensitive (case preserving, but insensitive for
access). I brought in a changeset that deleted a file with a
particular case, and added one with another case. The deletion
occurred after the add, so the net effect was a file having been
modified, and then the new name deleted.
For the second problem, I would think having a defined order might
help that exact scenario. I don't know what unwanted effects this
might have. For example, would I be able to do that before the rsync
that occurs?
--
Dustin Sallings
|
Try Searching:
servers, voip, java, networking, microsoft ...
|
|
|
|