Changes committed to git://oss.sgi.com/pcp/pcp.git dev
man/man1/dbpmda.1 | 2
man/man1/pmlogextract.1 | 2
man/man1/pmlogger.1 | 2
qa/269 | 16 +++
qa/src/test_pmcc.python | 162 +++++++++++++++++++++-------------------
qa/src/wrap_int.c | 13 +++
src/pmlogextract/GNUmakefile | 3
src/pmlogger/GNUmakefile | 3
src/python/pcp/pmapi.py | 34 ++++----
src/python/pcp/pmcc.py | 174 +++++++++++++++++--------------------------
10 files changed, 208 insertions(+), 203 deletions(-)
commit d54962a824c76e75f98d5354500e13d9e322fe59
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Wed May 29 10:38:58 2013 +1000
First updates to pmcc testing code, pylint clean too
commit 0c47a9cc9c054591b8e09869f4e9065384acbd47
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Wed May 29 10:38:29 2013 +1000
Gets the python convenience classes module up and stumbling
Fixes a number of out-of-date parts - particularly around the
handling of pcp.pmapi return values which nowadays use proper
exceptions for the error code. Lots of c-wrapper issues also
sorted out here though.
Additionally, changes the way the code was pushing timestamps
around to every single group, instead of just keeping this in
one spot in the surrounding group.
Plenty of work remains in this module, however - needs a fair
bit of work to make it pylint clean, testing, etc.
commit c325467afe9f60ced15f40bc59ebecd2871d042a
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Wed May 29 09:58:14 2013 +1000
Make pmlogger and pmlogextract first class citizens
Some discussion on IRC ended up with the what-a-good-idea
suggestion of making a couple of very important PCP tools
available on the default path. Previously for historical
reasons (I think) they were squirrelled away in BINADM_DIR.
This was the second time in as many weeks that people were
seeking logextract functionality and didn't know it already
existed cos of its unusual location.
This commit moves the two tools into /usr/bin and installs
back-compat symlinks in the old location for anyone with
hard-coded paths (scripts, etc).
The man pages no longer refer to them in the old locations
anymore as well. Also, noticed an incorrect reference in
the dbpmda(1) man page to its location being in BINADM_DIR
when in fact it is not - fixed.
This resolves Fedora bug 967709.
commit 4a4ed562c3ea0ea3919ddd6d50fbd0896cb26074
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Tue May 28 10:16:09 2013 +1000
Make test 269 reliable for all gcc version/optimisation levels
We no longer rely on the output of signed integer overfly from
qa/src/wrap_int.c as this produces undefined behaviour in modern
gcc versions. Could have tackled it via -fno-strict-overflow or
-fwrapv (gcc specific) options but these are also dependent on
gcc version and apparently are also buggy in some situations.
Further details: https://patchwork.kernel.org/patch/34925/
http://thiemonagel.de/2010/01/signed-integer-overflow/
|