[review] gdb: fix overload resolution for see-through references

["Tankut Baris Aktemur (Code Review)" <gerrit@gnutoolchain-gerrit.osci.io>]

Tankut Baris Aktemur has posted comments on this change.

Change URL: https://gnutoolchain-gerrit.osci.io/r/c/binutils-gdb/+/617
......................................................................


Patch Set 1:

> One question about the 'an_overloaded_function (const foo &&)' overload.  It would never be legal for the compiler to call this overload when doing:
> 
>  struct foo f;
>  an_overloaded_function (f);
> 
> So why does GDB even consider it when searching for a possible overload to call?  Let's say there exists the overload 'an_overloaded_function (const foo &&)' but the overload 'an_overloaded_function (const foo &)' does not exist, and the user calls 'an_overloaded_function (f)'.  Should GDB call the && overload, or should GDB say that no compatible overload was found?

I believe it's expected that GDB will search the function during overload resolution.  For an illegal case like this, I'd expect that it ranks the function with a very high badness value, so that an infcall is refused, just like the compiler rejects it.

Currently, if you give 'f' a const type, GDB makes the call (but it's still illegal).  If non-const like above, GDB says "Attempt to take address of value not located in memory."  It seems this is a separate bug; GDB resolved the function and attempted to call it but a failure occurred during call preparation.


-- 
Gerrit-Project: binutils-gdb
Gerrit-Branch: master
Gerrit-Change-Id: I39ae6505ab85ad0bd21915368c82540ceeb3aae9
Gerrit-Change-Number: 617
Gerrit-PatchSet: 1
Gerrit-Owner: Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
Gerrit-Reviewer: Keith Seitz <keiths@redhat.com>
Gerrit-Reviewer: Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
Gerrit-Reviewer: Tom Tromey <tromey@sourceware.org>
Gerrit-CC: Simon Marchi <simon.marchi@polymtl.ca>
Gerrit-Comment-Date: Fri, 29 Nov 2019 18:57:38 +0000
Gerrit-HasComments: No
Gerrit-Has-Labels: No
Gerrit-MessageType: comment