[clang] [clang][Diagnostics] Highlight code snippets (PR #66514)

[Timm Baeder via cfe-commits]

Timm =?utf-8?q?Bäder?= <tbaeder at redhat.com>,
Timm =?utf-8?q?Bäder?= <tbaeder at redhat.com>,
Timm =?utf-8?q?Bäder?= <tbaeder at redhat.com>,
Timm =?utf-8?q?Bäder?= <tbaeder at redhat.com>
Message-ID:
In-Reply-To: <llvm/llvm-project/pull/66514/clang at github.com>


tbaederr wrote:

> > One of the remaining tests fails because `PP->getIdentifierInfo()` ends up doing external lookup via `ASTReader`, which ends up emitting a diagnostic itself. Not sure what to do about that atm.
> 
> Can we just detect the reentrancy into the highlighting code and skip computing the highlight information entirely in that case, or is the problem more insidious than that? (Is the fact that diagnostic emission triggers lazy identifier deserialization itself a problem?)

The problem only happens if the deserialization causes emission of another diagnostic. We're already emitting a diagnostic (that's why we're highlighting the code), so that doesn't work.

One solution that seems to work is to just return an empty vector if `PP->getIdentifierTable().getExternalIdentifierLookup()` returns non-`nullptr`. That fixes the test case and my local highlighting still works. I'm just not sure if a) that disables highlighting for all PCH cases and if that's something people do a lot and b) if there might be other cases where identifier lookup might cause emission of a diagnostic.


https://github.com/llvm/llvm-project/pull/66514