Attention is currently required from: pespin.

neels has posted comments on this change. ( https://gerrit.osmocom.org/c/osmo-msc/+/37272?usp=email )

Change subject: do not invoke two Assignments (fixup for re-assignment) ......................................................................

Patch Set 1:

(1 comment)

Patchset:

PS1:

This patch may actually be more clear if you spit it into 2 steps (no need to really the change the […]

i would have added just a bool to indicate whether an assignment is pending. But that would introduce a potentially lingering blockage of all future Assignments, if it stays set "forever". So here I add a bool with a timeout.

I think we can't really split this patch, because the first half of the patch would add that plain bool without a timeout, which I don't want to have that way because it is "dangerous".

Adding a timeout to all Assignments is just a side-effect...

(I wish we had a proper osmo_fsm for assignment like in osmo-bsc, but introducing one now seems a lot of effort for "cosmetic" code changes, maybe avalanching to a huge refactoring... We should do that at some point, because we probably still have a bunch of corner cases now -- but it seems now is not the time to start that. I have too many patches pending for SDP, fmtp and AMR rates. Instead I want this "quick fix" because the re-assignment feature is already merged, and I uncovered this problem only now; so far the ttcn3 test was always just succeeding for me, and now it suddenly flipped over to uncover the race condition. My bad...)