[darcs-users] Two bug reports: patch bundle corruption and a missing error message

Martin Sjögren msjogren at gmail.com
Wed Jul 28 11:06:02 UTC 2004

On Wed, 28 Jul 2004 06:49:58 -0400, David Roundy
<droundy at abridgegame.org> wrote:
> Hmmm.  This is why the hash was added, since I had been told that mailers
> can do this (although I've never experienced it myself, althought that may
> be because I avoid CRLF).  I'd say it's definitely one of the mailers on
> the way which is responsible.

Yeah I tend to avoid CRLF myself, too, I have no idea why this
particular file used CRLF.

> Do you have any suggestion how the error messages could be improved to make
> it easier to guess what happened? Perhaps I should add to the "Patch bundle
> failed patch!" error message a note that this is most likely because a
> mailer has been creative with the line endings.

That would be helpful, I think. If the primary purpose of the hash is
to ensure that broken mailers haven't tampered with the contents, then
that should probably be reflected in the error message. :-)

> It's possible that when I fix the currently malformed MIME attaching code
> (it used to work) that mailers will be less likely to mess with the line
> endings.  Or maybe not.

We'll see. I'd be happy to test the code when it's finished, since I
seem to have a broken mailer somewhere along the path as a teset case.

> > 2. When I used darcs send --output=/tmp/foo, I got the unfriendly message
> >
> > Fail: takeLock: unsupported operation (Invalid cross-device link)
> >
> > (my /home and /tmp are on different hard drives). It kinda sucks that
> > I can't use /tmp for --output, but at least it would be better if the
> > error was caught and a nicer error message was printed... :)
> It looks like this is already fixed in the latest darcs (and indeed in
> darcs 0.9.23).  I'm not sure what the problem was (although I was able to
> reproduce it), but I know that Juliusz fixed some locking bugs, so
> presumably this is one of them.  I know there was a fix (which I didn't
> really understand) which fixed some sort of a problem when the _darcs
> directory was on a different file system from the working directory, so
> perhaps that was what fixed this.

Ah, great.


More information about the darcs-users mailing list