Re: Keyword DLM crashes with IDL 5.2?

[Randall Skelton <rhskelto(at)atm.ox.ac.uk>]

I think I have figured out the answer to my own question...

replace: IDL_KWGetParams(argc,inargv,argk,kw_err,NULL,1)
with:    IDL_KWGetParams(argc,inargv,argk,kw_err,argv,1)
                                                 ^^^^
or any other valid IDL_VPTR pointer array.

From the EDG...

  Plain_args is defined as an array of IDL_VPTR into which the
  IDL_VPTRs of the positional arguments will be copied.  This array must
  have enough elements to hold the maximum possible number of positional
  arguments as defined in IDL_SYSFUN_DEF2...

Reading the above suggests I need a finite array dimension as I am
expecting to pass 2 plain arguments.  The root of my problem was in the
EDG Example code they pass a NULL and I assumed I should do the same.  It
is somewhat surprising that it worked in IDL 5.3 and I didn't notice any
bizarre behavior...

Cheers,
Randall

On Mon, 30 Apr 2001, Randall Skelton wrote:

> Date: Mon, 30 Apr 2001 17:29:21 +0100
> From: Randall Skelton <rhskelto@atm.ox.ac.uk>
> Newsgroups: comp.lang.idl-pvwave
> Subject: Keyword DLM crashes with IDL 5.2?
> 
> Hi all,
> 
> I am trying to port a DLM that works in IDL 5.4/5.3 to work with IDL 5.2.  
> As a demonstration of my problem I have modified the file testmodule.c to
> include a simple procedure called 'testr' that takes a keyword and
> distributed it below.  It compiles under unix with:
> 
> gcc -Wall -Wimplicit -fpic -shared -I/usr/local/PACK/idl-5.3/external
> testmodule.c -o testmodule.so
> 
> Execution in IDL is:
> 
> IDL> testr, 1, a, ERROR=b
> % Loaded DLM: TESTMODULE.
> IDL> print, a, b
>           10           0
> 
> This routine compiles and works just fine under IDL 5.3, but it seg-faults
> during execution with IDL 5.2.  Without the keywords, the procedure works
> fine, and I can trace the fault to calling 'IDL_KWGetParams' in IDL 5.2.  
> I've read the relevant sections of the EDG at least twice over and after
> an afternoon of fiddling, I cannot find the bug.  Any and all help is
> greatly appreciated!
> 
> Randall
>