too many open raster files

[Martin Ameskamp]

> Any grass programmers still left out there?

        Sure.

> I get "WARNING: too many open raster files" after about
> 20 iterations of a dynamic simulation model that uses
> r.mapcalc and other GRASS programs, as well as some
> contributed code.  If I omit the following subroutine in
> our contributed code, then the program does not produce this
> error, so this is the offending code, although other things
> going on elsewhere could be contributing other open
> raster files (?):

        There are two aspects to this problem (I think). The first is
        the origin of the warning. This warning is produced in
        file src/libes/gis/opencell.c if you try to open more than
        MAXFILES files. MAXFILES is a constant that is #defined to be
        30 in src/libes/gis/G.h

        I don't know why it's 30, probably the programmers thought that
        nobody in their right mind would need more than 30 open files
        anyawy (sounds vaguely familiar...).
        So I guess you could change this constant and recompile GRASS
        (you'd probably need to recompile or at least relink just about
        all of GRASS since this is a fairly central point).

        The second aspect is that your use of more than 30 open files
        _may_ point to a resource leakage. On the other hand, a complex
        mapcalc calculation could easily lead to a large number
        of temporary files (I don't know what mapcalc does in main-memory
        and where it uses temp files; a look at the code might help there).

        If your program is a complex one that opens and closes raster files
        all over the place, a debugger might help. Set breakpoints in
        the relevant routines (G_open_cell etc., recompile them with
        debugging information if necessary) and check if close
        is called often enough etc.

        Finally, I only use G_gisinit at the beginning of my GRASS programs
        (mostly in main()). Calling it in a subroutine makes sense if that
        routine is the only one that uses GRASS code, but calling it
        more than once might have funny effects on some bookkeeping
        details, maybe including file descriptors.

        Good luck,  Martin
> 
> void probascii();
> 
>   {
> 
>     int fd;
>     char *mapset;
>     static char prog_name[] = "probascii";
>     static char filename[] = "DISTPROB";
> 
>     G_gisinit(prog_name);
> 
>     if((mapset = G_find_cell(filename,"")) == NULL) {
>       printf("Unable to find file %s.\n",filename);
>       exit(1);
>     }
> 
>     if((fd = G_open_cell_old(filename, mapset)) < 0) {
>       printf("Unable to open file %s.\n",filename);
>       exit(-1);
>     }
> 
>    ...CODE OMITTED
> 
>     G_close_cell(fd);
> 
>   }
> 
> I didn't write this code, but I'm trying to debug it.
> Anyone see why this fails to close the raster files?
> Or, any other clues as to what else could be
> producing this WARNING?
> 
> Bill Baker
> BAKERWL at UWYO.EDU
> 
> 
> 
> 


--
Martin Ameskamp, Inst. f. Informatik I (Computing Dept.)
Kiel University, Olshausenstr. 40, 24118 Kiel, Germany
Fax: ++49 431 8804054, Voice: ++49 431 8804474, 
ma at informatik.uni-kiel.de, URL http://www.informatik.uni-kiel.de/~ma/