Re: [Tinycc-devel] Small patch

[grischka]

Thomas Preud'homme wrote:
> > Maybe we should just put bcheck.o into libtcc1.a (and thus make via
> > the rules of lib/Makefile).  Unless there is a reason why we
> > shouldn't.
>
> Yes but that doesn't change the general assumption that we are compiling tcc 
> with gcc. If --cc is specified at configure time, then CFLAGS and CPPFLAGS 
> should be set when running make (make CPPFLAGS=foo CFLAGS=bar)

Sure, but if you want to solve the --cc=tcc/clang problem then the
question with using what compiler for libtcc1.a needs to be answered,
and also whether it makes sense to treat bcheck.o separately.

> Well, the test at the top of the function suggested me name could be NULL. I 
> suppose it can't be NULL when the symbol is global or common. What about a 
> malformed elf file?

Are you sure that putting an hash link to an not-existing name entry
is the right way to deal with a malformed elf file?

> > >              boolean.t = VT_BOOL;
> > >
> > > +            boolean.ref = 0;
> > Also redundant because type.ref is used with type.t = VT_FUNC only.
> >
> > I'm not completely opposed to make this look more solid but then
> > such patch should address the problem systematically such that the
> > code becomes smaller, not bigger.
>
> Can you develop about what you have in mind?

No, I just suspect.  There is mostly always a better way than
to "copy & paste" equivalent statements into several places, the
more if the statements are altogether logically redundant.

> > Same with
> >         pstrcpy(buf, sizeof buf, "a.out");
> > Because "a.out" cannot overflow buf[1024].
>
> sprintf could be used  then to save some space

I pushed some changes to that.

--- grischka