carbon-language/carbon-lang
1
0
Fork 0
mirror of https://github.com/carbon-language/carbon-lang.git synced 2026-10-05 22:02:55 +01:00
Code Issues Packages Projects Releases 10 Wiki Activity

Simplify freeing with new LLVM (#5334)

Browse Source 
We now have upstream support for a clean way to forcibly enable freeing
memory in a library context, so use that.
This commit is contained in:
Chandler Carruth
2025-04-18 22:05:25 +00:00
committed by GitHub
parent 89c9714825
commit 2bdea71c25
1 changed files with 2 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
toolchain/driver/clang_runner.cpp
+2 -8
View File
@@ -113,15 +113,9 @@ auto ClangRunner::Run(llvm::ArrayRef<llvm::StringRef> args) -> bool {
// execution here, as subprocesses leaking memory won't impact this process.
auto cc1_main = [enable_leaking = enable_leaking_](
llvm::SmallVectorImpl<const char*>& cc1_args) -> int {
// Clang doesn't expose any option to disable injecting `-disable-free` into
// the CC1 invocation, or any way to append a flag that undoes it. So to
// avoid leaks, we need to edit the `cc1_args` here and remove
// `-disable-free`.
//
// TODO: We should see if upstream would be open to some configuration hook
// to suppress this when it is generating the CC1 flags and use that.
if (!enable_leaking) {
llvm::erase(cc1_args, llvm::StringRef("-disable-free"));
// Last-flag wins, so this forcibly re-enables freeing memory.
cc1_args.push_back("-no-disable-free");
}
// cc1_args[0] will be the `clang_path` so we don't need the prepend arg.