Re: Unwarranted old-style cast warning and non-critical error

[Daniel J Sebald]

On 08/23/2012 05:04 PM, John W. Eaton wrote:
> On 23-Aug-2012, Daniel J Sebald wrote:
>
> | Most of these are what look to be one time C code, stored as an ll file,
> | converted to a cc file.  In such files it looks like casting 0 to
> | something is the common warning.  I'd help rid some of these, if you
> | like.  You'd have to review though.
>
> I don't know how you can get rid of warnings in code generated by flex
> unless you fix flex so that it conditionally generates C++ style casts
> when the output file is supposed to be C++ code.

Oh, flex creates that?  I didn't know it added that much.  Can some 
redefining macros be used somehow to compile that file?  I.e., define 
DBL_MAX to be some valid C++ cast?  Or is it better to go after a fix 
for flex?


> | There are a few easy ones in there like "unused args".
>
> Some of those need to actually be fixed by making the code do
> something useful, not simply by removing the symbol names from the
> argument lists.

Leave them as warnings then.  Or touch the variables in some trivial way.


> | Otherwise, there is some (probably original from years back) C++ code
> | syntactical issues in one file.  This actually may be more of a concern
> | than one would think because I see "conflicts: 16 shift/reduce, 1
> | reduce/reduce".
>
> That's a message from Bison about the Matlab/Octave language grammar,
> not a C++ syntax problem.
>
> | Also there are some shadowed variables which doesn't look like a major
> | issue, just that "index" must be declared globally somewhere.  I'm not
> | seeing it anywhere; most uses of "index" appear to be member variables
> | or member functions.  And the code that the warnings are coming from is
> | C code so it can't be shadowing a member variable...warning says
> | "global" anyway so again, not member.  Is there a function or variable
> | with name "index" in some library header?
>
> "index" is a C function.
>
> |<strings.h>  has a function "index".  Perhaps that is the conflict.
> | Commenting out that header file from gl2ps.c results in those warnings
> | going away (so yes, it is a conflict with<strings.h>) but of course
> | then memcpy, strlen, strcpy aren't declared and create other warnings.
> |
> | So that is it.  I've attached a patch for gl2ps.c which replaces "index"
> | with "idx" in the offending locations if you want.  ("idx" didn't appear
> | in gl2ps.c prior to this patch.)
>
> I don't want to fork and maintain gl2ps.c separately from its upstream
> source, so patching it in the Octave sources is not what we want to
> do.  Instead, get the problems fixed upstream and then we can import a
> new version into the Octave sources.

Oh, that originates from somewhere else?  I'll drop the developer(s) a note.

Most of the low hanging warnings fruit has been picked then.

Dan