test: improve SILGen coverage on Windows #21371
Conversation
Make these SILGen tests pass on windows too
|
@swift-ci please test |
|
CC: @gottesmm, @eeckstein |
|
CC: @slavapestov @jrose-apple - at the suggestion of @gottesmm |
| @@ -1,9 +1,9 @@ | |||
|
|
|||
| // RUN: %target-swift-emit-silgen -module-name witness_tables -I %S/Inputs -enable-source-import %s -disable-objc-attr-requires-foundation-module -enable-objc-interop > %t.sil | |||
| // RUN: %target-swift-emit-silgen -emit-sorted-sil -module-name witness_tables -I %S/Inputs -enable-source-import %s -disable-objc-attr-requires-foundation-module -enable-objc-interop > %t.sil | |||
There was a problem hiding this comment.
Why the change to sorting here? (and elsewhere, but especially this file)
There was a problem hiding this comment.
The order differs between Linux/Darwin and Windows. I don't think that there is a way to get both sides to work without the sorting. Perhaps there is some other approach that I didn't consider.
There was a problem hiding this comment.
Hm. I'm pretty sure the order is supposed to be based on source, so that's a little suspicious.
There was a problem hiding this comment.
My fear about this is that the need for it could show a source of non-determinism.
There was a problem hiding this comment.
@jrose-apple, empirical observation indicates that this is not the case. There is certainly a possibility of there being a bug in SILGen.
|
I want this to land. @compnerd can you split out the -emit-sorted-sil changes. I do think it is important to figure out /why/ this is happening. I talked with Jordan and we think the rest of it is ok. |
|
Splintered the unsorted bits into #21493 |
|
Going to close this, however, this is still unfixed. The change in #22308 fixes the sorting issue. |
Make these SILGen tests pass on windows too
Replace this paragraph with a description of your changes and rationale. Provide links to external references/discussions if appropriate.
Resolves SR-NNNN.