This is mandatory. You can't ever not do that. In fact, I consider
it best practice to do the following:
I've run out of time to finish this, but, the attached patch starts to
bzero(3) all malloc(3) calls. I know this is overkill, but for the
time being, it's better to be safe than sorry. I'm alarmed at a few of
the calls in dbmail_imap_session_new(), however. Why on earth are we
malloc(3)'ing a char? The pointer is at least four bytes (depending
on your architecture), but the resulting allocation is only one byte.
Why bother at all? Why not just make self->tag a character instead of
a pointer?
self->tag = (char *)my_malloc(sizeof(char));
self->command = (char *)my_malloc(sizeof(char));
self->args = (char **)my_malloc(sizeof(char **));
?? This really baffled me. Other notes, sizeof(char) is always going
to be 1, no matter what. These calls are being optimized away and it's
rather silly to have them in the code. Lastly, memset(,0,) is not as
fast as bzero(3). I've moved several calls from memset(,0,) to
bzero(3) and would suggest the rest of the tree be swept for similar
changes. Anyway, I was about halfway through dbmail-imapsession.c
before I ran out of time with this. I'll post the patch in a few hours
once I finish sweeping through the tree. -sc
--
Sean Chittenden
