On Friday 19 September 2008 16:44:22 Daniel Jacobowitz wrote:
> It looks like you don't have a copyright assignment or employer
> disclaimer on file.  Is that correct?  If so, we should sort it out
> now before you contribute anything larger.

Working on that already. Might take a while, though.

The stuff in question is certainly well below any reasonable threshold
of "original art", though, but I see the problem.

> > I hope this is uncontroversial, but before going further into this
> > direction I have two related questions:
> > 
> > 1. How would the prefered way to call, say,  xfree on a 'conceptionally const
> > char *' item look like? Are casts to non-const (void *) acceptable
> > here?
> 
> Probably acceptable, but if you know it is allocated by xfree, should
> it really be returned as const?

The problem will come up with  "char *"  struct members pointing to
"names" of all sorts. They could arguably be "const char *" as they
usually aren't touched between construction and destruction to
get a compiler warning if something tries to modify them nevertheless.
But then, those members could stay non-const...

> > the work. Are there any shortcuts available/acceptable?
> 
> If you have to update the call sites, you can say "all callers
> updated" or something like that in the changelog entry, though if any
> of the changes are more than mechanical then it's better to list
> the complicated ones explicitly.  But for just a change to a prototype
> and a definition, I think the changelog entry is still appropriate.
> 
> BTW, your changelog is too verbose in another way; there is normally
> only one starred entry for a file and you can combine similar changes.
> Like this:
> 
> 	* symfile.c (allocate_symtab, allocate_psymtab, free_named_symtabs)
> 	(deduce_language_from_filename): Accept a const argument.

Ok.
 
> Please use diff -up for patches; with the default diff arguments, it's
> impossible to see what you've changed.

I intented to do that but I somehow fumbled. Next try attached.

Andre'