On 04/06/15 09:28, Nathan Scott wrote:
----- Original Message -----
Reviewed this one from Nathan and check -g pmie passes on a couple of QA
platforms, so this is good to go.
Thanks Ken!
Since no good deed should go unpunished - could you cast an eyeball or
or two over the commit below as well? (either now or later, no rush - I
am pretty sure it's nice, simple & correct, but since its pmlogger...)
This one is already in my tree (no cherry pick needed).
The man page documents the new -o (bit of a bugger that -L was already
used!) and -K ... but the QA test uses --local-PMDA and no -K option ...
the former probably does not matter, I'll extend 948 to exercise the latter.
-o implies -L ... is that intended? If there is nothing to log, why not
exit unless -L is specified, like for other "host" contexts?
In pmlogger's fetch.c I think the call to __pmEncodeResult in myFetch()
should have kept the first parameter as ctxp->c_pmcd->pc_fd not 0 ...
this is never called in the PM_CONTEXT_LOCAL case.
Now the first argument to pmEncodeResult() is not used these days, but
might be used again in the future if we have a new version dependency in
the protocol (which is what the "fd" argument is used for).
I've reverted this bit of the change in my tree and it passes all -g
pmlogger QA (including the modified 948).
And finally the pmnewlog usage message was wrong on multiple fronts, but
that was mostly prior damage, not part of the most recent commit ...
I've fixed that too.
So I have commits ready to be pushed for all of these ... just need your
feedback and an answer on the -o => -L question.
|