Fix singleton operations #2909
Fix singleton operations #2909
Conversation
(cherry picked from commit a1e8e58)
…many there are (cherry picked from commit 93e7384)
so it fixes a memory leak on finalize Signed-off-by: Gilles Gouaillardet <[email protected]> (cherry picked from commit fb5bcc4)
Signed-off-by: Ralph Castain <[email protected]>
|
yeah, the rest of the problem with singleton comm_spawn is in the orte layer - i'll address it tomorrow |
|
@hppritcha @jsquyres I would suggest going ahead with this PR and #2897 for now. With the changes in this PR, singletons will work just fine with dstore enabled or disabled. Singleton comm_spawn hangs, but that is a secondary problem and can be dealt with later. I don't have time right now, so I'd suggest someone file an issue for v2.1 and we'll deal with it when time is available. |
|
@hppritcha Signed-off-by checker failed because @rhc54 cherry-picked some master commits from before we required signed-off-by. They're commits from @ggouaillardet, who a) is a proven community member, and b) has a signed contribution form on file for commits from before we required signed-off-by. My $0.02: good to go (hopefully Travis will finish soon). |
|
Note I have a pr in a similar signature sign off issue. Do we give founding members who commit directly to master without sign offs a pass? Note that isn't the case with this PR. |
|
This isn't a question of getting a "pass". All we are doing is porting a commit already in our code base over to a release branch. There is no code being introduced into the repository, and so technically no sign-off is required. New code requires a sign-off, regardless of who is committing it. |
|
👍 on @rhc54's answer. |
Cherry-pick several commits from master to fix singleton operations. Includes some partial custom repair of singleton comm_spawn - still some problems that need to be resolved.
Refs #2897
Signed-off-by: Ralph Castain [email protected]