On Thu, 26 Jan 2006, David Berry wrote:
>> Just tried it out on an ADAM task (to see about leaks) and it's
>> essentially useless because on exit from the action the locator count
>> invariably goes up because the parameter system does not annul its
>> locators. So you simply find out that you got 3 new parameter locators
>> during the action...my enthusiasm is waning.
>
> Presumably this only happens on the first pass through the application (if
> you are running from orac-dr or from icl, etc). Once the parameters have
> been created the hds locator count should presumably settle down. Could
> you just skip the test on the first pass through the application?
The new and improved HDS_INFOI (hdsInfoI - yes, a capital 'I' but that is
what HDS seems to be going towards) is now available.
hdsInfoi(NULL, "LOCATORS", "!KAPPA_MON,!STATS", &n, status );
will now return the number of active hds locators excluding any that start
with KAPPA_MON or STATS. (probably should include a "." in that string
above to be sure of not matching STATISTICS).
I've even patched kappa_mon so that it now reports leaks.
One subtlety is that in kappa_mon I need to mask "!HDS_SCRATCH" as well
since it seems that HDS_SCRATCH files get created and not cleaned.
With this in place, Brad's test case with ndf sections now does report the
leak immediately.
--
Tim Jenness
JAC software
http://www.jach.hawaii.edu/~timj
|