Changes committed to git://oss.sgi.com/pcp/pcp.git dev
debian/changelog | 5
man/man1/pmdabash.1 | 4
src/pmdas/bash/event.c | 2
src/python/pcp.py | 485 ++++++++++---------------------------------------
4 files changed, 111 insertions(+), 385 deletions(-)
commit acdc96573393bbbd68ea66f25e5bd3ad7262031f
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Thu Oct 4 20:02:01 2012 +1000
Remove now-redundant locking layer from Python PMAPI wrapper
We currently have two levels of locking for Python clients,
due to the original python wrapper predating the threading
work done in libpcp. This change removes all locking from
the pcp.py layer (both the BigPythonLock & lastUsedContext).
In doing this, several dormant bugs were exposed (and fixed
at the same time), some quite subtle - happily the code is
much simplified, so these stood out:
- pmLookupDesc (error code not checked within the loop)
- pmSetMode (incorrect status checking, spurious zeroing)
- pmRecordSetup (missing unlock on error path)
- pmRecordAddHost (missing unlock on error path x 2)
I *think* (unrelated) issues remain in the following calls:
pmErrStr_r, pmUnitsStr_r, pmIDStr_r, pmInDomStr_r
(the calls do not match the C calling convention - can this
be right? surely not? not sure though, maybe some Python
magic happens under the covers here).
commit a8550fe00e17b8b0ef43c64c76a302799b24e3a5
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Thu Oct 4 19:46:55 2012 +1000
Make a note that pmdabash is experimental
commit ab3cf824e6e8a15e51eca3e64ff8eec30520fe10
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Thu Oct 4 19:45:39 2012 +1000
Bump version in debian changelog too, missed on previous
|