This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: RFC: MI - Detecting change of string contents with variable objects
On Friday 22 December 2006 10:12, Nick Roberts wrote:
> > > Because -var-evaluate-expression uses varobj_get_value so they will always
> > > return the same value?
> >
> > Then, there are two solutions:
>
> We seem to be agree that the patch does the right thing and are just talking
> about implementation details.
Yes.
>
> > 1. Make c_value_of_variable and friends accept struct value as opposed to
> > taking struct varobj.
>
> c_value_of_variable, as it's name implies, requires a struct varobj as it's
> argument.
You can renamed it to anything you want, say "c_format_value", or whatever.
> > 2. Extra the part of c_value_of_variable that you've based your function on
> > into a separate function. Make both c_value_of_variable and install_new_value
> > call the new function.
>
> value_get_print_value is quite a small function. After you've wrapped
> some statements in if clauses and worked out how to call them with a
> common argument, I can't see that you would gain much.
>
> > I'm not sure which approach you find better, but I don't think copy-pasting
> > is a solution to anything.
>
> Clearly it is *a* solution, it's just a question of whether it's the *best*
> solution in this case. Anyway, it's not a straight copy-paste (the asserts are
> removed, for example).
Probably it's perfectionism, but if we want -var-update to return a varobj if the
result of -var-evaluate-expression would be different, then we must execute
the same code as for -var-evaluate-expression when deciding if the value has changed.
If you run some other code, however similar *now*, it's very likely that soon things
will break.
> I think if you look through the code you'll find
> numerous examples where one section of code is a slight variation of another.
Yes. I'd like to remove all such cases.
- Volodya