[clangd] Enable parsing of forwarding functions in the preamble by default

[HighCommander4]

Fixes clangd/clangd#2324

[llvmbot]

@llvm/pr-subscribers-clangd

@llvm/pr-subscribers-clang-tools-extra

Author: Nathan Ridge (HighCommander4)

Changes

Fixes clangd/clangd#2324

---

Full diff: https://github.com/llvm/llvm-project/pull/127359.diff

1 Files Affected:

• (modified) clang-tools-extra/clangd/Compiler.h (+1-1)

diff --git a/clang-tools-extra/clangd/Compiler.h b/clang-tools-extra/clangd/Compiler.h
index 4e68da7610ca2..e513e4c40794a 100644
--- a/clang-tools-extra/clangd/Compiler.h
+++ b/clang-tools-extra/clangd/Compiler.h
@@ -40,7 +40,7 @@ class IgnoreDiagnostics : public DiagnosticConsumer {
 
 // Options to run clang e.g. when parsing AST.
 struct ParseOptions {
-  bool PreambleParseForwardingFunctions = false;
+  bool PreambleParseForwardingFunctions = true;
 
   bool ImportInsertions = false;
 };

[upsj]

Maybe for consistency we should also change the other defaults that IIRC are used to configure tests:

llvm-project/clang-tools-extra/clangd/ClangdServer.h

Line 187 in dbc98cf

bool PreambleParseForwardingFunctions = false;

llvm-project/clang-tools-extra/clangd/ClangdServer.h

Line 504 in dbc98cf

bool PreambleParseForwardingFunctions = false;

[HighCommander4]

Maybe for consistency we should also change the other defaults that IIRC are used to configure tests:

llvm-project/clang-tools-extra/clangd/ClangdServer.h

Line 187 in dbc98cf

bool PreambleParseForwardingFunctions = false;

llvm-project/clang-tools-extra/clangd/ClangdServer.h

Line 504 in dbc98cf

bool PreambleParseForwardingFunctions = false;

Done

[upsj]

LGTM!

[hokein]

The change looks good. We should be mindful of the potential performance regression due to this (1% increase in preamble size, as noted https://reviews.llvm.org/D124688#3483895).

I’ll leave the final approval to @kadircet.

[kadircet]

thanks, as long as we're keeping the underlying option I think this makes sense! we can flip it in production if we see big regressions

[kadircet]

thanks, as long as we're keeping the underlying option I think this makes sense! we can flip it in production if we see big regressions