[darcs-users] darcs on MS Windows (was: internal error)

[David Roundy]

On Sat, Dec 06, 2003 at 08:59:33AM -0800, Kevin Smith wrote:
> David Roundy wrote:
> >To use ssh or scp, you'd need ssh installed (but I doubt anyone would want
> >to use ssh who doesn't already have it).
> 
> Makes sense. That affects parts of push and push-and-apply, right? 
> Anything else?

Pull and get, if you want to pull or get via scp.

> >.zip is already in binary.  I don't know what .class and .jar files are.
> >If they are like .o files in C, then I'll add them to boring, but they are
> >something you might actually want in the repo, I'll add them to binary.
> 
> .class files are the Java equivalent of .o files. I can imagine a 
> project wanting to check in some .class files, but it should be rare. So 
> I think boring is the right choice.
>
> .jar files are something like a .so or .lib file. They are generally a 
> collection of .class files plus other stuff, such as resources. Projects 
> would often want to check in third-party jars, so these should probably 
> be binary instead of boring.

Ok.

> > Currently darcs looks at the file contents during the record to guess
> > if it is binary, so there shouldn't be a problem (the slowness you saw
> > when you added large binary files was almost certainly because of the
> > very poor binary to hex conversion darcs had (until very recently).
> 
> I guess I'm still bothered that the default behavior will (always) be to 
> add some files that I probably don't want added.

Hmmm.  I guess I tend to see it as the user's business to add only the
files they want to add.  But my view isn't necesarily the best one to go
by, as it tends to be skewed by what darcs historically was capable of.  In
this case both recursive adds (and --look-for-adds) and the boring filter
are pretty new (I haven't created a "real life" repository since they were
added), I tend to think of them as "extra" features.

> Perhaps in some future version, darcs could have a list of file types 
> that should be individually confirmed before being added. That way, you 
> could still do a recursive add on a new project, but would be likely to 
> end up with the results that you really wanted.

Something like this could be added, although I'm hesitant to try to make
darcs "know" too much about how users should be handling their
repositories.  It's better to give it the flexibility to allow users to
choose for themselves what they want it to do.

A subset of this idea might be to add an interactive flag (and possibly gui
as well) to add, which would prompt the user for each file.  Then they
could approximate your feature idea by running:

darcs add --recursive .
darcs add --boring --recursive --interactive .

This would prompt for every boring file, rather than a subset of boring
files, but at least wouldn't prompt for all the interesting files.  This
shouldn't be too bad if they have a decent "make clean" they could run
prior to doing the darcs adds.

> I have been hoping to upgrade to a newer darcs for a while, but keep 
> getting spooked by reports of new problems, so I have been waiting for 
> the real 0.9.15 release. Now I'm very interested to experiment with its 
> new binary detection behavior.

Well, 0.9.15 will be released by the end of this week, as I'll be moving
next Monday, and definitely want to get the release done by then.  I just
need to polish off a few bugs reported by a coworker and figure out why the
cgi script is broken again.  After 0.9.15 is out, I'll finally be able to
start getting rid of all the old patch formats, which'll be nice.
-- 
David Roundy
http://www.abridgegame.org