Re: [PATCH] Use std::optional when available

[Pedro Alves <pedro@palves.net>]

On 2023-10-10 17:49, Tom Tromey wrote:
>>>>>> "Andrew" == Andrew Burgess <aburgess@redhat.com> writes:
> 
> Andrew> I think this change needs to be held off until we move to C++17 - at
> Andrew> which point gdb::optional can just be dropped.
> 
> Once we switch I think we can just remove this code entirely.

Agreed, once we can rely on std::optional, we should delete all the tests
under gdb/unittest/optional/.  They were originally copied from libstdc++'s
testsuite, to make sure our gdb::optional behaved the same as std::optional,
in the subset of features it did support.  See d35d19584cf5.

BTW, with C++17, gdb::string_view can be replaced with std::string_view too,
and the whole of gdb/unittests/basic_string_view/ can go away too.

Pedro Alves

> 
> GDB 14 branched, so I think we're ready for that now.
> 
>>> +#if __cplusplus < 201703L
> 
> Andrew> Ignoring my previous comment for a moment; we now rely on this version
> Andrew> number being defined the same in two files.  Would it be better to move
> Andrew> the '#define GDB_OPTIONAL' into gdb_optional.h?  Then we can be certain
> Andrew> that it will be defined correctly, and in sync with any future changes
> Andrew> made in that file?
> 
> I don't think that would really help.  IIUC the issue would be compiling
> two source files with different -std settings.  But in this case
> gdb_optional.h would react to this.  Maybe I'm not understanding what
> you mean though.
> 
> It seems to me that we simply don't need to support compiling the tree
> with different std settings for different files -- it should be fine to
> just assume that everything is compiled compatibly.
> 
> Tom