Re: feature freeze? (Re: [Announce] mutt-1.5.13 is out)

[Rocco Rutte <pdmef@xxxxxxx>]

Hi,

* Matthew D. Fuller [06-08-17 00:34:44 -0500] wrote:

> Oh, much more than a 1-liner (which was part of the argument).
> Different functions get called for gdbm vs bdb etc, so it had to be
> repeated in a number of places.  It could probably be factored down to
> something a little cleaner (a few possibilities occur to me just
> glancing at it), but since it works for me.

Ouch, but sorry, the patch can be done much simpler and much more 
efficient (calling functions like strdup() always involves 
malloc()/free() which should be avoided as much as possible because 
they're horribly expensive).

The way to do it IMHO is this: in mutt_hcache_per_folder() you only need 
to change the way the MD5 sum is computed to strip the trailig /. The 
reason is that we only want to get rid of it for the MD5-based database  
filename so that a folder with and without the trailing slash gets has 
the MD5 sum; we really want to ignore everything else for fixing this 
problem. There, you can do this:

  char tmp[_POSIX_PATH_MAX];

  strfcpy(tmp,folder,sizeof(tmp));
  size_t tmplen=mutt_strlen(tmp);
  if (tmp[tmplen-1]=='/') tmp[--templen]=0;

  MD5Update(&md5, tmp, tmplen);

(untested, of course) and that should be pretty much it.

(The one-liner idea was stupid as we shouldn't mangle the folder 
variable's contents in there)

Opinions?

  bye, Rocco
--
:wq!