[PATCH] D118921: [Format] Don't derive pointers right based on space before method ref-qualifiers

[Ben Hamilton via Phabricator via cfe-commits]

benhamilton requested changes to this revision.
benhamilton added inline comments.
This revision now requires changes to proceed.


================
Comment at: clang/lib/Format/Format.cpp:1949
           continue;
+        // Don't treat space in `void foo() &&` as evidence.
+        if (const auto *Prev = Tok->getPreviousNonComment()) {
----------------
Do we also need to worry about `T && foo();` style declarations?



================
Comment at: clang/lib/Format/Format.cpp:1951
+        if (const auto *Prev = Tok->getPreviousNonComment()) {
+          if (Prev->is(tok::r_paren) && Prev->MatchingParen)
+            if (const auto *Func = Prev->MatchingParen->getPreviousNonComment())
----------------
If this is parsing a general function declaration, I think it's not safe to assume that the token immediately before the ref-spec is the close-paren.

It looks like `Prev` can be a few other things, including at least:

* `tok::kw_const`
* `tok::kw_volatile`
* `tok::kw_throw`
* `tok::kw_noexcept`

and probably C++ attribute(s) as well:

> noptr-declarator ( parameter-list ) cv(optional) ref(optional) except(optional) attr(optional)	

* https://en.cppreference.com/w/cpp/language/function
* https://en.cppreference.com/w/cpp/language/declarations

I *think* they can also be in any order, so the ref-spec could come at the very end after cv-qualifier, an expection specification, and C++ attributes.



Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D118921/new/

https://reviews.llvm.org/D118921