Mark,
> Not sure who is the ARY expert round here...
RFWS??
> I recently upped the limits on the number of NDFs usable in a single
> CCDPACK task from 100 to 1000 in response to a user who had a lot
> of frames to reduce at once. He now sees this:
>
> !! All 512 slots allocated for entries in the ARY_ facility Data Control Block
> ! have been used up.
> ! ARY_FIND: Error finding an array in an HDS structure.
> ! NDF_OPEN: Error opening an NDF data structure.
> ! Unable to get an NDF identifier for 'CB222a-KS-524_ss'
> ! WCSEDIT: Modification of WCS components failed.
> ! Application exit status ARY__EXSLT, exceeded number of allocated slots
>
> A quick look through the ARY source shows this is running into the
> ARY__MXDCB = 512 limit, defined in ARY_CONST. I can bump this up
> (might as well slap the same multiplier - 4? - on the other limits
> defined in that file too) and supply the user with a new version
> easily enough. It would seem like sense to do the same to the values
> in NDF_CONST as well.
>
> (a) does anyone know any reason that this would be inadvisable?
Can't think of any. Rodney would probably have documented any issues
concerning the size of these arrays, if there were any issues.
>
> (b) beyond NDF are there other library packages where changes are likely
> to be required?
NDG_CONST currently implies a limit of 100 on the number of NDF identifier
it can issue, but I think in fact this value is not used any more (i.e.
there is now no limit in NDG). I'll remove the appropriate constant
definition from NDG_CONST.
>
> (c) do we want to check the increased values into the CVS repository
> for future use?
Can't see why not (if it works!).
David
|