[OpenSIPS-Devel] [OpenSIPS/opensips] 1bbe7a: Fix un-released transaction upon t_wait_for_new_br...

[Bogdan Andrei IANCU]

  Branch: refs/heads/2.4
  Home:   https://github.com/OpenSIPS/opensips
  Commit: 1bbe7ab88271f58916cd60debe63f728f5598e82
      https://github.com/OpenSIPS/opensips/commit/1bbe7ab88271f58916cd60debe63f728f5598e82
  Author: Bogdan-Andrei Iancu <bogdan at opensips.org>
  Date:   2020-04-13 (Mon, 13 Apr 2020)

  Changed paths:
    M modules/tm/t_reply.c

  Log Message:
  -----------
  Fix un-released transaction upon t_wait_for_new_branches + cancel.

If t_wait_for_new_branches() is used and no real signaling branch is added, the caller canceling must not be ignored upon branch selection. The injected 408 timeout on the phony branch must be considered, leading the transaction termination.

Thanks to Chris Maciejewski for reporting and providing troubleshooting info.
Also thanks to Razvan Crainea ( @razvancrainea ) for investigating and reproducing the problem.

(cherry picked from commit 7221b984d555cb68fd07e1bfb502607446345d42)


  Commit: ef3ffb0f3890c824026439a6e7b8ec28093f291e
      https://github.com/OpenSIPS/opensips/commit/ef3ffb0f3890c824026439a6e7b8ec28093f291e
  Author: Bogdan-Andrei Iancu <bogdan at opensips.org>
  Date:   2020-04-13 (Mon, 13 Apr 2020)

  Changed paths:
    M modules/tm/t_fwd.c

  Log Message:
  -----------
  Improve canceling of transactions doing only t_wait_for_new_branches().

If a transaction has no real signalling branch, but only doing t_wait_for_new_branches(), upon caller cancel inject a 487 reply on the phony/waiting branch in order to terminate the transaction on the spot.
The provious behavior was to wait (even after the cancel) for the phony/waiting branch timeout (fr_inv_timeout) and to generate an 408 timeout.
Now, there is no useless waiting and a proper 487 is generated.

(cherry picked from commit 8af80a40bb9877b91380da670353a68c53f0a36b)


Compare: https://github.com/OpenSIPS/opensips/compare/2ee910093066...ef3ffb0f3890