Re: [PATCH] git-mailsplit: with maildirs try to process new/ if cur/ is empty

[Alex Riesen <raa.lkml@gmail.com>]

Karl Hasselström, Tue, Nov 06, 2007 17:35:48 +0100:
> On 2007-11-06 15:51:09 +0000, Johannes Schindelin wrote:
> > On Tue, 6 Nov 2007, Jeff King wrote:
> > > On Tue, Nov 06, 2007 at 11:01:03AM +0000, Johannes Schindelin wrote:
> > > > I fail to see how the absence of one of cur/ or new/ can lead to
> > > > the absence of patches. You could forget to save some patches,
> > > > yes, but the presence of cur/ and new/ is no indicator for that.
> > >
> > > Read my message again. Alex is proposing ignoring errors in
> > > opening the directories; I am proposing ignoring such errors
> > > _only_ when the error is that the directory does not exist.
> > >
> > > IOW, if there is some other error in opening the directory, it
> > > should be fatal, because you might be missing patches.
> >
> > Yeah, sorry, I missed that.
> 
> I think it might actually not be totally unreasonable to error out
> unless both directories exist. From
> http://www.qmail.org/qmail-manual-html/man5/maildir.html:
> 
>   A directory in maildir format has three subdirectories, all on the
>   same filesystem: tmp, new, and cur.
> 
> In other words, if it doesn't have these three directories, it isn't a
> Maildir directory.

On the same line of reasoning, if opening a ".../cur" fails with ENOTDIR,
it must be not a Maildir...

> On the other hand, one could argue that requiring both dirs to exist
> is being too picky.

...which MUST NOT mean it does not contain useful patches.
IOW, the tool can try and apply everything it finds.
If user told it to get patches from the...whatever, then the patches
should it get and damn qmail.