Re: [PATCH] Check record types for compatibility in ada_type_match

[Tom Tromey <tom@tromey.com>]

>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:

>> +/* Helper for ada_type_match that checks that two record types
>> are
>> +   compatible.  As with that function, FTYPE is the formal type and
>> +   ATYPE is the actual type.
>> +
>> +   Note that it is ok if this function is not precise, as long as
>> +   there aren't too many false negatives.  That is, it's better to
>> +   return 'true', because that will result in a menu being presented
>> +   to the user.  */
>> +

Tom> I don't fully understand the reasoning here.  AFAIU, if there is only
Tom> one match, no menu will be presented.  In that case, is returning true
Tom> still a conservative choice?

Returning true means that gdb will consider the types compatible.

If there are multiple overloads of some function, ada-lang.c checks them
for compatibility with the actual arguments; see ada_resolve_function.
If multiple functions then survive this compatibility check, gdb will
present a menu.

So, if there's some doubt, returning true will mean that more possible
overloads will be considered matches; and in this case a menu will be
presented.

Mainly here we are interested in filtering out the types that definitely
do not match.  IME it's annoying when debugging Ada to be given
nonsensical choices.  I'd like to even further refine this code, but Ada
is pretty complicated and so it's not easy to do.

Tom