diff options
| author | Matthias Krüger <matthias.krueger@famsik.de> | 2025-02-04 18:49:39 +0100 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-02-04 18:49:39 +0100 |
| commit | 46b18a91abc7fc2fb902c0157c594f5ec3701e02 (patch) | |
| tree | c93c011c393c66b952c95a59036ebe61f6429aae /compiler/rustc_codegen_gcc/src | |
| parent | b07fa7696b87da65b09d022ff4c1211361ab0ee4 (diff) | |
| parent | d2a30f7d884fecfea60ed540aa830a75013e43b8 (diff) | |
| download | rust-46b18a91abc7fc2fb902c0157c594f5ec3701e02.tar.gz rust-46b18a91abc7fc2fb902c0157c594f5ec3701e02.zip | |
Rollup merge of #136477 - lqd:nll-tls-spans, r=matthewjasper
Fix a couple NLL TLS spans Some NLL TLS tests show incorrect spans for the end of function. It seems that the `TerminatorKind::Return` source info span can sometimes point at the single character after the end of the function. Completely changing the span where the terminator is built also changes a bunch of diagnostics: small functions have more code shown unrelated to the errors at hand, wrapping symbols appear and weird-looking arrows point to the end of function, etc. So it seems this is somehow unexpectedly relied upon in making diagnostics look better and their heuristics. So I just changed it where it matters for these few tests: the diagnostics specialized to conflict errors with thread locals. r? `@matthewjasper`
Diffstat (limited to 'compiler/rustc_codegen_gcc/src')
0 files changed, 0 insertions, 0 deletions
