Skip to content

Conversation

jsquyres
Copy link
Member

@jsquyres jsquyres commented Mar 2, 2017

Two fixes from @bosilca on master:

  1. Do not free() a buffer on the stack.
  2. Fix a code THREAD_MULTIPLE bug in the TCP BTL (per MPI_Barrier hangs in 10-20h long runs #3042)

@bosilca I had to fix a conflict when porting over the #3042 fix. It seems to work, but I'd appreciate your review.

@hppritcha I marked this as for v2.1.0 because it fixes a real bug that a user encountered (thanks to Evgueni Petrov @Welcome-Dear-Evgueni for reporting).

@jsquyres jsquyres added the bug label Mar 2, 2017
@jsquyres jsquyres added this to the v2.1.0 milestone Mar 2, 2017
@jsquyres jsquyres requested a review from bosilca March 2, 2017 00:06
bosilca and others added 3 commits March 1, 2017 16:30
This is mostly for error cases, where we need to release the
newly created proc. Currently the code deadlocks because the endpoint
lock is help at the release and the lock is not recursive.

Aslo added some code to print the IP addresses that don't match during
the TCP connection step.

Thanks to Evgueni Petrov @Welcome-Dear-Evgueni for reporting the
issue.

Signed-off-by: George Bosilca <[email protected]>

(cherry picked from commit ec4a235)

Signed-off-by: Jeff Squyres <[email protected]>
Signed-off-by: George Bosilca <[email protected]>
(cherry picked from commit b0f8d2c)
Make the message more friendly / more detailed, and de-duplicate it
(just in case it happens a lot).

Signed-off-by: Jeff Squyres <[email protected]>
(cherry picked from commit 5b484c9)
@jsquyres jsquyres force-pushed the pr/v2.1.x/bosilica-fixes branch from a149dae to a0d29ac Compare March 2, 2017 00:33
@jsquyres jsquyres mentioned this pull request Mar 2, 2017
@jsquyres
Copy link
Member Author

jsquyres commented Mar 2, 2017

@hppritcha I think this PR is worthy of v2.1.0; we can discuss and see if you agree.

It is not vital to get into tonight's MTT, but is still fixing some real bugs, and paving the way to get more information to solve #3035.

@hppritcha hppritcha merged commit ddc8e48 into open-mpi:v2.x Mar 2, 2017
@jsquyres jsquyres deleted the pr/v2.1.x/bosilica-fixes branch March 2, 2017 14:37
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants