[Asterisk-Dev] Voicemail password changes not sticking

[James Armstrong]

I finally tracked down a problem we have been having, users have not 
been able to change their voicemail passwords and have them stay after a 
reload.

The password would be changed in 'realtime', but when the voicemail.conf 
file would be re-written, the changes would not be put in. It turned out 
to be something I have not been able to explain looking at he C code, 
because it looks ok, but here is what was happening:

When reading lines from the voicemail.conf file to match on the Context, 
the context name it was parsing would be missing the last digit so the 
user would never match and therefore the new password would never be 
written. For example, if our context is [harksystems], the parser would 
end up with the variable 'currcontext' set to 'harksystem'. The code 
looked ok, but I was able to fix it by changing the math of the tmpctx 
and tmpctxend when parsing the context from the []'s. Here is the code 
from vm_change_password:

                         if (inbuf[0] != '\0') { /* skip over parsing 
for lines starting with a comment character and empty lines */
                                 /* Check for a context, first '[' to 
first ']' */
                                 tmpctx = strchr(inbuf, '[');
                                 if (tmpctx) {
                                         tmpctxend = strchr(inbuf, ']');
                                         if (tmpctxend && (tmpctxend > 
tmpctx)) {
                                                 /* Valid context */
PROBLEM IS WITH THE FOLLOWING TWO LINES: 

    ast_copy_string(currcontext, tmpctx + 1, tmpctxend - tmpctx-1);
    currcontext[tmpctxend - tmpctx - 1] = '\0';
IT WORKS IF I CHANGE TO:
    ast_copy_string(currcontext, tmpctx + 1, tmpctxend - tmpctx);
    currcontext[tmpctxend - tmpctx] = '\0';

                                         } else {
                                                 tmpctx = NULL;
                                         }
                                 }

Like I said, it looks like the tmpctxend-tmpctx-1 is correct, but it 
chops off the last character in the context when it is parsed from 
between the []'s.

Thanks,
James