[asterisk-dev] [Code Review] 2742: res_musiconhold cleanup

[wdoekes]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/2742/
-----------------------------------------------------------

Review request for Asterisk Developers and Tilghman Lesher.


Bugs: ASTERISK-21775 and ASTERISK-22252
    https://issues.asterisk.org/jira/browse/ASTERISK-21775
    https://issues.asterisk.org/jira/browse/ASTERISK-22252


Repository: Asterisk


Description
-------

This review concerns cleanup of the res_musiconhold code.

(1) Removed the special REF_DEBUG code which -- as far as I could tell -- only caused problems when enabling REF_DEBUG.
    @tilghman: if the code should still be in there, could you explain why?
    See bug: ASTERISK-22252

(2) Reduce duplicate code, remove unreachable code, remove commented-out code.
   
    Among other things:
    - Removing the HANDLE_REF argument to moh_register(). Doing the unref manually
      takes only one line of code.
    - init_app_class() gets a bit less code, but is now reused by another caller. I
      had to add "XXX: why don't we check for failure (ast_timer_open()) here?"
      because I didn't get why there was a difference.
    - Renamed `mohclass` to `other` in moh_register(). The module won't get an A for
      properly named variables.

(3) Use `init_files_class` instead of `moh_scan_files` directly, because it returns
    <0 for failure and >0 for success and it was only checked for zero (= no files).

(4) Added the patch by Michael Young from ASTERISK-21775, which seems correct to me.

(5) free()'s were encountered that should be ast_free()'s.
    The astobj2.c:162 INTERNAL_OBJ: bad magic number 0xdeaddead for object 0x7faf44978818
    errors I get are not solved, but they're still here when noloading musiconhold,
    so they're unrelated.

(6) Added comment about LOGIC by tilghman:
    - This referred to the previous situation the bare class pointer was compared
      against the state->class pointer. This was changed to take state->name into
      account.
    - When we get to this part, class is reffed, so comparing anything to class should
      be ok, even if state->class is not reffed. Right?

(7) Fixed so realtime cachedrtclasses works with non-files mode.


Diffs
-----

  /trunk/res/res_musiconhold.c 396165 

Diff: https://reviewboard.asterisk.org/r/2742/diff/


Testing
-------

Used the refcounter(1) to check that things were fine when:
- moh reload'ing
- playing standard musiconhold ("files")
- playing cached RT musiconhold ("files")
- playing uncached RT musiconhold ("files")
- playing standard musiconhold ("quietmp3")
- playing cached RT musiconhold ("quietmp3")
- playing uncached RT musiconhold ("quietmp3")

Tested that the quietmp3 mode re-used the same spawned mpg123 for both realtime MOH classes when CACHERTCLASSES is true.


Thanks,

wdoekes

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20130805/292b17a3/attachment.htm>