Minor bug in graph from plotutils-2.4.1

[Emile Talbot]

Hi,

I have managed to compile plotutils-2.4.1 on DJGPP using the GCC 3.3.3
compiler and binutils 2.15.  However, I had trouble getting "graph" to
work even though the compilation process had not detected any errors.
For testing purposes, I used an example from the plotutils documentation
with different output formats.  I did notice that "graph" did work for
me once or twice but, more often than not, it would fail to give a valid
output file.  I tried compiling it under C++ to catch type-mismatches
and again with the compile option -Wuninitialized, all without success.

>From an observation of the portable metafile output, I guessed that a
parameter relating to the y-scaling was not being initialized correctly.
After a bit more detective work, I noticed that the  variable
"multigrapher->y_trans.output_min" was not being initialized to zero in
plotter.c where it definitely looked like it should be
("multigrapher->x_trans.output_min" was being initialized to zero twice
instead).  That is, in the plotter.c file I have, line 1049 should be 
  multigrapher->y_trans.output_min = 0.0;
not
  multigrapher->x_trans.output_min = 0.0;

With that modification, graph.exe seems to be working as expected.  

On a different note, the DJGPP port of GCC defines the type of the first
argument of the iostream.write method to be (const char*) instead of
(const unsigned char*).  Furthermore, the DJGPP port of GCC defines the
type of the fsync function (defined in <unistd.h>) to be int.  Source
files
    c_defplot.cc
    g_flushpl.cc
    g_write.cc
    i_rle.cc
    n_write.cc
    p_defplot.cc
    z_write.cc
all had to be modified slightly to cast these arguments into the type
expected by the GCC C/C++ libraries to compile correctly under C++ (for
libplotter.a).

Thanks for these utilities.  I'll have to examine them more closely to
see if these plotting utilities will truly be useful to me.  Certainly,
I have made good use of the GNU Compiler Collection.  Again, thanks.

I do have a question, though.  How, if in any way, are these GNU
plotutils related to the earlier GNUplot?

Emile Talbot
(address@hidden)
(address@hidden)