With these changes, the QA group "sanity" passes for the most part (with
only a couple of odd failures on a couple of platforms) on all hosts in
the QA Farm.
Changes committed to git://oss.sgi.com/kenj/pcp.git dev
configure.in | 49 ++++++++++++++++++++++++++++++++-------
man/.gitignore | 1
src/include/platform_header.h.in | 2 +
src/libpcp/src/logutil.c | 12 +++++++--
src/pmdas/sample/src/sample.c | 22 +++++++++++++++--
src/pmdas/solaris/disk.c | 1
src/pmieconf/rules.c | 18 ++++++++++----
7 files changed, 87 insertions(+), 18 deletions(-)
commit cfddd47be335622277ffb0b697db0109b482b4f0
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Aug 24 14:40:52 2011 +1000
configure.in - another refinement to the typedef and printf format stuff
commit a7de37098a97e417a97c32fb887a2d63985a6b6d
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Aug 24 11:07:41 2011 +1000
sample PMDA - Solaris sysinfo() is different
same name, API is totally different ... deal with it.
commit 247913432a90477129c40601f6f05f29a9a40557
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Aug 24 07:37:57 2011 +1000
More compilation warning removal - trivial cases
commit f7e3e5c78fad76e29d2898f4ce55ed81edd630dc
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Aug 24 07:29:24 2011 +1000
pmieconf/rules.c - another tricky compiler warning case
Following the example from logutil.c in libpcp, the problem appears to
be the (apparently) bogus __attribute__((warn_unused_result)) marking
of strtol() and strtoul() in glibc and gcc's inflexibility to over-ride
the __attribute__ marking.
In this case we don't care about the value from strtol() and strtoul()
we're interested in updating the pointer to get past the integer value.
The gross fix that appears to work on all platforms is to declare a
dummy variable to receive the return value, and then mark that variable
as unused using the ugly __attribute__((unused)) construct.
We were already locked into gcc, but that decision is now set in concrete!
commit a09304afac94be6eef15fffea5590ba496715a1f
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Aug 24 07:25:09 2011 +1000
configure.in - tweak printf format guessing logic
For Mac OS X, a pthread_t is really a pointer to a struct ... need to
make the cpp post-processing smarter, and add a case to emit "p" (as
in %p) as the appropriate format when the unchained typedefs end up at
a pointer.
commit 258b071d7bf163f97f8a7552b9263b28da6a6c15
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Tue Aug 23 20:44:32 2011 +1000
logutil.c - trial code change to remove compiler warnings
Hard case around strtol() and (apparently) bogus __attribute__
marking in glibc, and gcc inflexibility to over-ride the
__attribute__ marking.
|