Re: [bug-gnubatch] Compiler warnings (as of 1.6 rel)

[Jan Schampera]

On 09.09.2012 14:32, John Collins (Xi Software Ltd) wrote:

> > --> blindly use socklen_t?
> The trouble about doing that is that it falls over on some platforms. If
> someone can suggest a better check in the configure.ac script I'll
> change it.
>
> Linus has some strong comments about the confusion with versions!

Sorry, not expirienced enough.

Shouldn't it be enough to compile a snip that uses the header and 
socklen_t, and define a macro expanding to int or socklent_t using the 
result of that compile (actually I thought that's what the configure 
test already does)?


> > btcharge.c:35: warning: ‘Filename’ defined but not used
> > --> Might be about (not) using ABORT_NOMEM, which is true in some
> > other places, too (which just are not compiled, I guess).
> Oh yes the should come out. It doesn't use anything which allocates
> memory any more. Likewise other things.

Spring clean :-)


> > y.tab.c: In function ‘yyparse’:
> > y.tab.c:1392: warning: implicit declaration of function ‘yylex’
> > --> predeclare yylex in the atcover parser generator?
> That's a quirk-ette of yacc/bison libraries I think.

Yeah, that's a known one, AFAIR.


> > gbatch_jobdata.c: In function ‘gbatch_jobdata’:
> > gbatch_jobdata.c:115: warning: implicit declaration of function ‘sigset’
> > --> probably something wrong with HAVE_SIGSET?
> That's me being lazy about setting up ignoring signals in a portable
> way. I think that sigset should be defined in signal.h if the library
> supports it.

The header mentiones it, guarded by some directives. That's why I 
thought that the configure script might go wrong somewhere.

--
Be conservative in what you do, be liberal in what you accept from others.
- jbp, master of the net, in RFC793