Re: [Gm2] Numerical Library : GM2 issue with multidimensional array parameter

[Gaius Mulley]

Michael Riedl <address@hidden> writes:

> Gaius,
>
> I got the latest stable version (6.4 branch) of GM2 compiled on a 64
> bit Linux system - so I now began with the porting activities.
>
> I am making some progress on the two libraries - the IO Modules gets
> compiled and is partially working. The module TestReal (check NAN &
> INF) is a stub now - need to fix that sooner or later as I would like
> to catch a quite NAN in some cases ...
>
> Concerning the numerical part I try to get rid of some issues caused
> by the non-standard handling of EXPORT / IMPORT clauses in local
> MODULES in the XDS compiler ... the solution will most probably be
> some limited number of preprocessor directives.
>
> Another issue are compiler directives. The ISO standard defined them
> as being surrounded by <* somedirective *> without any more details if
> I am not totally mistaken. GM2 regards that as an error so I have to
> clean them out - think I'll get that fixed. Or is there an option to
> tell the compiler to simply ignore them ?
>
> At the moment the main issue I now see may be illustrated by the
> attached file - GM2 is complaining about using parts of a
> 2-dimensional open array as a 1-dimensional argument to another
> routine.
>
> I cannot find something being wrong with the code so I would like to
> ask you to have a view on that.
>
> Thanks in advance
>
> Michael

Hi Michael,

I've looked at the open array - it certainly appears to be a bug - will
fix it.  Thanks for the report.  I'll also modify gm2 to consume
compiler directives - initially and then decide how these should be
handled later on.  But prior to these changes I'll backport the range
test and NaN detection code into 6.4.0 - hopefully this weekend,

regards,
Gaius