On 25/07/13 13:28, Nathan Scott wrote:
- 366 - presence of nfs2 and nfs3 client metrics is enabling
logging, ...
I think this one is pretty easy to resolve ... 366 is checking pmlogconf
migration (could be retired altogether one day), so
as long as _some_ metrics are being expanded the test is useful.
I've just pushed a commit to fix the nfs2/nfs3 case and document in the
test how to maintain it going forward.
- 491 - test failure: failed to use consistent hash key?
This one is failing on two different big-endian/64-bit
platform/architectures and never fails on x86 for me.
Does this ring any bells? The src/keycache2.c hash key
scheme seems to be unintentionally platform specific,
somehow.
[77%] 491 - output mismatch (see 491.out.bad)
3,7c3,7
< keys 104649 & 121066 hash to 1647692332
< key-104649 -> 1647692332
< key-121066 -> 1282769657
< keys 62832 & 82985 hash to 1557163883
< key-62832 -> 1557163883
---
keys 104648 & 121065 hash to 1647692332
Unfortunately it is not my hash() implementation, but this is really
strange ... key -> hash one one platform and key+1 -> same hash on
another platform, but only for some keys. Maybe need big endian or
little endian variant 491.out selection (see qa/122 for how this could
be tested for).
I don't have access to a big endian 64-bit machine, so can't offer more
help I'm afraid.
|