Re: [RFC] Revisit PR 16253 ("Attempt to use a type name...")

[Keith Seitz <keiths@redhat.com>]

On 06/16/2015 10:54 AM, Doug Evans wrote:
>
> This approach is an improvement with no ill effects (that I can see),
> so I'm ok with it.
> Please add a reference to PR 16253 in the "hack" comment
> in the code.
> 

Will do.

> Do other callers of symbol_matches_domain need similar treatment?
> I was wondering about block_lookup_symbol.

Yeah, I wonder about that myself, but I am cautious to commit to that
unless I can coverage test it. So far, I've not been able to trigger a
failure in any of these.

The proposed change is (intentionally) very conservative. While I could
be convinced to add it (nearly?) wherever symbol_matches_domain is used,
coverage testing will be difficult, if not impossible. [I spent many,
many hours doing coverage testing of the completer API change. It was
gruesome!]

> btw, here's some perf data using the gmonster1-pervasive-typedef.exp
> test from my monster testcase generator.

Thanks for that! I haven't yet had a go at it this.

Keith