wagnerdm at seas.upenn.edu
wagnerdm at seas.upenn.edu
Fri Nov 16 17:19:56 CET 2012
Seems I need to defend myself a bit, so here goes.
Quoting Gwern Branwen <gwern0 at gmail.com>:
> On Fri, Nov 16, 2012 at 10:40 AM, <wagnerdm at seas.upenn.edu> wrote:
>> 1. Patches made for the darcs-1 format can't be applied after the upgrade. I
>> think it's fairly likely that there are people out there who maintain their
>> own patches or who are currently working on something and merely haven't
>> sent in their patches. I don't want to make their lives harder unless
>> there's a good reason to.
> 2.0 format became default years and years ago, and xmonad development
> has been stable/stagnant for as long. They can upgrade their own repo
> if they want to send a patch from however long ago (which seems
> unlikely since it's bad practice to keep a fork private for a long
While it's true that xmonad-core has not had a lot of activity
recently, that's not equally true of xmonad-contrib (which is the
repository in question). Additionally, your claim about what other
people can do after upgrading the central repository doesn't mesh with
my reading of the darcs documentation. Here's the scenario as I see it:
1. Person A upgrades repository A into A-2
2. Person B upgrades repository B into B-local-2
2. Person B clones repository A-2 into B-2
It's not clear that this results in a state where person B can
transfer patches from repository B-local-2 into B-2. In particular,
"darcs help convert" says,
"Furthermore, darcs 2 repositories created by different invocations of
this command SHOULD NOT exchange patches, unless those repositories
had no patches in common when they were converted."
> Do you really think the Darcs
> team would've switched to 2.0 format, released it, set it as default,
> continue to release for years and years, and publicly recommend using
> it as default if those diffs were not either irrelevant or
I'm talking about repository metadata here, not patch metadata.
Repository metadata is not always copied during a convert or a clone.
I fully believe that darcs-2 is capable of storing all the same data;
just not that I, personally, know how to copy it from the old
repository to the new one.
And please, the adversarial tone of this note really seems
unwarranted. The darcs wiki itself recommends _not_ upgrading hashed
darcs-1 repositories to darcs-2 repositories.
>> 4. And finally, according to the wiki, the only advantage of darcs-2 is
>> better handling of conflicts, which I don't think have been a significant
>> problem for us. All the other advantages are already available in
>> hashed-format darcs-1 (which the -contrib repository already has been
>> upgraded to).
> For long-lived private branches, better handling of conflicts is
> important... So which is it, #1 or #4? Are there long-lived private
> patches per #1 in which case #4 is wrong, or are there not, in which
> case #4 is right but #1 wrong?
Is it really so clear that there will be lots of conflicts from
long-lived private patches...? I don't believe these things are really
so mutually exclusive as you say.
More information about the xmonad