I made some changes over the last two weeks ( 2fc6557bd4 and
8f1d3ed0d6a2) that should allow a command line length of up to 1944
characters.
We've checked it doesn't break anything it at EAO by running some
longish ORAC-DR recipes.
David
On 31 January 2017 at 14:48, Malcolm J. Currie <[log in to unmask]> wrote:
> David,
>
>> Does anyone have thoughts on the safety or otherwise of simply raising
>> the value of MESSYS__VAL_LEN to - say - 2000 ? Or is it just a case
>> of trawling through all its usages and considering each one?
>
>
> This is an old chesnut. FOr instance it was discussed in 2006 and what
> other parameters would need enlarging beside or other fixed limits that
> might prevent us having very long command lines.
>
> The proposal then was for Brad to make and test the change at JAC with
> ORAC-DR. If nothing breaks, commit the change. I think 1000-characters was
> the suggestion, although Mark did suggest something like 4k to avoid coming
> back to this a few years down the line. When it was written memory was at a
> premium.
>
> Peter wrote:
>
> "David, looking in msp_par.h there's a definition of MSP_MXMSGSZ of 2036.
> That defines the content part of the structure sent between processes, so I
> think that is the actual limit, goodness knows how that relates to the 500
> byte message limit, but clearly it wouldn't be a good idea to exceed it.
> Increasing the 2036 would have performance issues as that is the size
> actually sent up and down the wire.David, looking in msp_par.h there's a
> definition of MSP_MXMSGSZ of 2036. That defines the content part of the
> structure sent between processes, so I think that is the actual limit,
> goodness knows how that relates to the 500 byte message limit, but clearly
> it wouldn't be a good idea to exceed it. Increasing the 2036 would have
> performance issues as that is the size actually sent up and down the wire."
>
> If there's aren't other limitations, let's increase the buffer size.
>
> Malcolm
|