Re: [unavailable regs/locals, 01/11] registers status upwards

[Tom Tromey <tromey@redhat.com>]

I read this thread today.  A bit late, I know, but since it impacts
post-7.3 work (I hope :-) I would like to bring it up.

Jan> In such case these functions should be `warn_unused_result' and
Jan> they should be renamed from the original functions.  The original
Jan> function names should throw an exception (for the case of infcall
Jan> etc.) - if any <unavailable> code should be handled directly by the
Jan> callers the original function names can even internal_error on
Jan> <unavailable> values.

If I read correctly, we're all in agreement about this as a long term
goal.

Jan> I wanted to express general disagreement with this style expecting
Jan> preciseness and no mistakes by the developers, which is not
Jan> considered to be a "safe enterprise development style".

I think we generally agree about this too.


My understanding is that the code in 7.3 is at least as robust as
gdb-before-the-series-was-applied.  That is, the changes were, if not
"fail-safe", at least "fail-compatibly".

Pedro, are you planning to implement the API changes post-7.3?
If not, could you file a bug report about it?

Tom