On 2004-02-08 17:24:52 -0500, David Yitzchak Cohen wrote:
> On Sun, Feb 08, 2004 at 04:44:35PM EST, Thomas Glanzmann wrote:
>
> > > Also, I think what I said in my previous mail isn't 100% correct: if you
> > > look at a new message, it should be moved from the new/ subdir to the cur/
> > > subdir, and if you delete it, its filename is changed, IIRC ... in both
> > > of the above cases, mtimes of the new/ and cur/ should be updated, and
> > > you may want to argue that no other changes really matter 99% of the time.
> > > (For the other 1% of the time (i.e., for functions like edit-type, etc.),
> > > you can force an immediate cache invalidation before exiting the function
> > > in question.)
> >
> > The problem is that I care about the 1% see [1]. Especially in the case
> > of an external modification but we could do it configurable. My patch
> > doesn't invalidate the cache on edit (maybe it does, but not explicit),
> > deletion or change flags.
>
> <quote from="[1]">
> Afte editing files in a maildir's cur or new subdir mutt does not notice
> the change and does not update hcache.db. Instead the messages are
> displayed with a currupted message body.
> </quote>
The message body corruption could be fixed by stating the message file
before or after opening it (which should be negligible overhead) and
discarding/updating the cache when the file is newer than the cache.
> Why don't you just make a special file in new/ and in cur/ and touch
> it every time you touch a message?
If you have to touch anything explicitely, why not the directory? That
way you will only have to stat the directory, instead of the directory
and the special file.
hp
--
_ | Peter J. Holzer | Shooting the users in the foot is bad.
|_|_) | Sysadmin WSR / LUGA | Giving them a gun isn't.
| | | hjp@xxxxxxxxx | -- Gordon Schumacher,
__/ | http://www.hjp.at/ | mozilla bug #84128
Attachment:
pgp7kT4gTSLjO.pgp
Description: PGP signature