----- Original Message -----
> Details for the most commonly failing tests are available from
> ftp://oss.sgi.com/projects/pcp/download/qa-status/
>
(this is great, thanks Ken).
> === Synopsis ===
>
> QA Status from Ken McDonell's QA Farm in Melbourne
>
> Summary: 23344 run, 31 failed (0.13%)
>
> QA Test Fail Failure Signatures (number of different failure patterns)
> 878 10 6 2 2
These appear to be race conditions - I've pushed through a handful of
changes to make things more reliable on slower hosts, let's see how
that goes.
> 802 5 5
This one's odd - appears to be new this week, but there's nothing I can
think of since the previous run that would have induced this (dbpmda is
seeing a timeout on PMDA start?) ... something in libpcp? Surely not.
This is the bad diff... (errors originating from dbpmda/src/pmda.c)
> Error: __pmGetPDU() failed: No such PMAPI error code (28684)
6a7,8
> __pmGetPDU(11): Timeout waiting for a response from PMCD
static void
pmdaversion(void)
{
int sts;
__pmPDU *ack;
int pinpdu;
pinpdu = sts = __pmGetPDU(infd, ANY_SIZE, _creds_timeout, &ack);
if (sts == PDU_CREDS) {
if ((sts = agent_creds(ack)) < 0) {
fprintf(stderr, "Warning: version exchange failed "
"for PMDA %s: %s\n", myPmdaName, pmErrStr(sts));
}
}
else {
if (sts < 0)
fprintf(stderr, "__pmGetPDU(%d): %s\n", infd, pmErrStr(sts));
else
fprintf(stderr, "pmdaversion: expecting PDU_CREDS, got PDU type
%d\n", sts);
(gdb) p /x 28684
$1 = 0x700c
grep 0x700c src/include/pcp/impl.h
#define PDU_CREDS 0x700c
M'kay - so thats a/ not negative (how did we get into that first
else branch?) and b/ *is* PDU_CREDS! (how did we not go into the
sts==PDU_CREDS code path in the first place?)
Confused. Am I missing something there? Compiler bug???
cheers.
--
Nathan
|