Unwarranted old-style cast warning and non-critical error

[John W. Eaton]

On 17-Aug-2012, Daniel J Sebald wrote:

| When building Octave, gcc issues this old-style cast warning:
| 
| In file included from builtins.cc:9:0:
| ../../octave/src/interpfcn/variables.h:105:40: warning: use of old-style 
| cast
| ../../octave/src/interpfcn/variables.h:105:65: warning: use of old-style 
| cast
| 
| It is due to the following line of code, and in particular the DBL_MIN 
| and DBL_MAX:
| 
| extern OCTINTERP_API octave_value
| set_internal_variable (double& var, const octave_value_list& args,
|                         int nargout, const char *nm,
|                         double minval = DBL_MIN, double maxval = DBL_MAX);
| 
| I think this may be a gcc compiler oversight.  I read somewhere that the 
| warning went away with gcc version 4.5.0-5, but I have version 4.5.1 
| (from late 2010) and I'm still getting the warning.
| 
| I tried to get rid of the warning with "static_cast<double>(DBL_MIN)" 
| but that did not help.
| 
| Can someone please confirm this warning disappeared somewhere along the 
| way in the last year and a half?
| 
| ...
| 
| This leads to a question.  Is the line above intended instead to be the 
| following (i.e., replace DBL_MIN with -DBL_MAX)?
| 
| extern OCTINTERP_API octave_value
| set_internal_variable (double& var, const octave_value_list& args,
|                         int nargout, const char *nm,
|                         double minval = -DBL_MAX, double maxval = DBL_MAX);
| 
| I'm assuming that the default definitions are meant to be that there is 
| no restriction on the value, as with the integer case defined before the 
| double case:
| 
| extern OCTINTERP_API octave_value
| set_internal_variable (int& var, const octave_value_list& args,
|                         int nargout, const char *nm,
|                         int minval = INT_MIN, int maxval = INT_MAX);
| 
| However, INT_MIN and DBL_MIN are not analogous.  DBL_MIN is the smallest 
| positive double value (i.e., greater than zero).  I printed out the value:
| 
| DBL_MIN = 2.22507e-308
| 
| If the minval is meant to be -DBL_MAX, apparently this doesn't make a 
| difference in behavior, otherwise a bug would have shown up long ago.

Right.  I went with -octave_Inf and octave_Inf instead.

I also checked in another change that replaces all uses of DBL_MIN,
DBL_MAX, DBL_EPSILON, FLT_MIN, FLT_MAX, and FLT_EPSILON in C++ code
with std::numeric_limits functions.

jwe