Fix usNIC CIDs #9992
Fix usNIC CIDs #9992
Conversation
|
@j-xiong How does one get feedback on the ci/cloudbees/stage/Summary and continuous-integration/intel/jenkins CI failures? Clicking the Details links seem to try to go to internal Intel URLs. |
|
@jsquyres Unfortunately only someone from Intel can see the details. For this PR, I checked the error yesterday and saw infrastructure related job cancellations and restarted the run. The new failure is a single transient mpichtestsuite failure with the tcp provider, which is a known issue and is unrelated to the PR. |
Cast time(NULL) down to 32 bits. For this specific case, the time is used as a sequence number. It is definitely expected that the query will complete within 2^31 additional querys down to libnl (if it doesn't, we have larger issues), so even if sizeof(time_t)==8, it's ok if we truncate it down to the lower 32 bits. Signed-off-by: Jeff Squyres <[email protected]>
Ensure that a data structure is fully zeroed out. Signed-off-by: Jeff Squyres <[email protected]>
Fix a potential thread race condition. Signed-off-by: Jeff Squyres <[email protected]>
This uninitialized value is ignored in this specific devcmd, but we might as well assign it to 0 for code readability. Signed-off-by: Jeff Squyres <[email protected]>
Ensure order of operations doesn't truncate a value: force the use of 64-bit math. Signed-off-by: Jeff Squyres <[email protected]>
jsquyres
force-pushed
the
pr/usnic-cid-fixes
branch
from
c128f4e to
5f78e36
Compare
|
Forgive me for being out of the libfabric loop for so long -- what's the merge policy these days? |
|
@jsquyres Generally we expect CI runs to pass and receive at least one approval before merge. However sometimes exceptions are made (e.g. totally unrelated CI failure). |
See individual commit messages for details.