================
@@ -3686,14 +3686,14 @@ Preprocessor::LexEmbedParameters(Token &CurTok, bool
ForHasEmbed) {
std::pair<tok::TokenKind, SourceLocation> Matches) {
Diag(CurTok, diag::err_expected) << Expected;
Diag(Matches.second, diag::note_matching) << Matches.first;
- if (CurTok.isNot(EndTokenKind))
----------------
a-tarasyuk wrote:
@shafik Thanks for the feedback. `LexEmbedParameters` is shared by `#embed` and
`__has_embed(...)`, so `EndTokenKind` is still required to terminate normal
parsing appropriately.
This change only affects error recovery. `DiscardUntilEndOfDirective` is
specifically intended to skip the remainder of the directive line, so once
`CurTok` is already `tok::eod`, there is nothing left to discard, and the
context-specific expected end token (e.g., `tok::r_paren` for `__has_embed`) is
no longer relevant.
The added tests cover this behavior. Other scenarios appear to be handled by
the existing `embed___has_embed_parsing_errors` tests; if there are additional
edge cases to cover, please let me know.
https://github.com/llvm/llvm-project/pull/175104
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
