[darcs-users] a patch and a bug

[John Meacham]

On Wed, Jul 30, 2003 at 07:28:49AM -0400, David Roundy wrote:
> On Tue, Jul 29, 2003 at 06:26:29PM -0700, John Meacham wrote:
> > attatched is a patch which fixes (somewhat) the mime encoding of outing
> > emails. There are still some issues remaining (which will require
> > changing sendMail to fix) but this should make them behave a bit nicer
> > at least..
> 
> I'll apply this after dealing with the push bug below.  On the subject of
> the encoding of patches, really the encoding is the same as whatever the
> files in the repo are, and I have no control over that, so it's a little
> hard to know what encoding to report...

yeah, I am working on a proper patch for sending MIME email, I wrote the
code before for another haskell project involving email bounce messages
and can reuse it almost exactly. This patch was mainly a stopgap measure
to keep things like half a quoted principle conversion happening to the
patch in transit (which would be bad-bad) and to have them appear in
peoples mailers in a somewhat prettier format.

Yeah, the character set of the patch data itself can't very well be
defined, as a single patch may contain text from several different files
with different encodings and perhaps even binary data. But since the
patch mime type is given as a text/*, many mailers may try to interpret
it as text, so I specified the charset that the patch meta info is
written in. (by which I mean everything except the added lines
themselves)


> 
> > also, darcs push keeps on claiming i have no recorded patches to push,
> > despite my creation of this patch.  (hence me sending this patch
> > directly to the list) Pull works just fine to pull patches from the
> > darcs repository.. I am not quite sure what I am doing wrong.
> 
> This troubles me... it sounds like a bug in get_common_and_uncommon, which
> is a very important function, but also sufficently complicated that it may
> very well have a bug.  If so I definitely want to find that bug!
> 
> Just to first check in case you did indeed make a mistake, did you try
> giving darcs push a repository argument?  i.e.
> 
> darcs push http://abridgegame.org/darcs
> 
> If not, give that a try.  darcs push (or pull) with no arguments works with
> the last pushed or pulled repository, so if you had two repositories of
> your own (one for experimenting, one for interfacing to ag.o), you might
> very well have pulled locally and then accidentally pushed locally to the
> repo you just pulled the patch from.
> 
> If this wasn't the problem, could you perhaps make a tarball of your
> repository available to me, so I could try to figure out what went wrong?
> Or if you could put it up on an http server, that would work.  Because my
> only guess as to where the bug might be lurking (since push has worked
> consistently for me) is that it could be triggered by an unusual layout of
> intentory and inventories/.

hmm.. okay, it appears the problem was that the last repo was set to '.'
for some reason. my _darcs/prefs/lastrepo had '.' and _darcs/prefs/repos
had both '.' and 'http://abridgegame.org/darcs'. curious. It would be
nice if push gave a nice summary of what it did, even with the -v option
it is not very chatty, basically a message saying 
sent:
patch1
patch2
patch3
to 'foo+darcs at wheretherepois.com' from 'john at foo.net' unsigned.
my previous patch added the last line, since it was non-trivial to
figure out where darcs push was sending my patches when i had no
_darcs/prefs/email set. it just succeeded without any messages.
        John


-- 
---------------------------------------------------------------------------
John Meacham - California Institute of Technology, Alum. - john at foo.net
---------------------------------------------------------------------------