On Wed, Feb 04, 2026 at 01:04:20PM +0800, Kevin J. McCarthy wrote: > This function is used by mx_get_magic() to check if a path is a > maildir mailbox. I believe it was only looking for "cur" to save on > stat calls. > > However, this can cause misbehavior in some cases. The browser won't > enter a (non-maildir) directory that contains only a subdirectory > "cur". On the other other hand, mutt relies on the "tmp" and "new" > subdirectories existing or fails operations, such as append to > mailbox, or postponing mail. > --- > Just going through some branches I'd accumulated over the past few years. > > I'm sure I don't need to ask for feedback, but this is a tradeoff of a > couple stat calls for correct behavior in some less common situations. > Thoughts?
This could plausibly break things for anyone who has written one or more old maildir directories to a read-only medium, and to save space has removed the empty 'new' and 'tmp' directories. On the other hand, implementing this change would discourange users from doing the same thing going forward, so overall I think it's the correct approach. Would it make sense to drop the check for 'new' and 'tmp' IFF 'cur' can only be opened read-only? Admittedly, anybody who removes 'tmp' and 'new' has already created a non-standard directory structure that other programs would not be guaranteed to recognise. But that doesn't mean that somebody out there hasn't done it. Maybe have the 'missing new and tmp, but on a read-only medium' case throw a warning, and see if anybody ever mentions seeing the warning. If not, remove the workaround after a couple of years.
