[OM] Fix finalization of nested ReferenceValue #8325
Conversation
See llvm/circt#8325 Signed-off-by: unlsycn <unlsycn@unlsycn.com>
See llvm/circt#8325 Signed-off-by: unlsycn <unlsycn@unlsycn.com>
See llvm/circt#8325 Signed-off-by: unlsycn <unlsycn@unlsycn.com>
See llvm/circt#8325 Signed-off-by: unlsycn <unlsycn@unlsycn.com>
|
cc @mikeurbach |
See llvm/circt#8325 Signed-off-by: unlsycn <unlsycn@unlsycn.com>
There was a problem hiding this comment.
I think this makes sense, but can you please include more info about what this is trying to resolve? This would be good info to have in the commit message when it merges. I think I can understand it from the test case, but in general, we request to have good commit messages: https://llvm.org/docs/DeveloperPolicy.html#commit-messages
@uenoku please take a look as well
See llvm/circt#8325 Signed-off-by: unlsycn <unlsycn@unlsycn.com>
4832257 to
e076e48
Compare
Sure! |
e076e48 to
63a16e2
Compare
After instantiating an Object in the Evaluator, we recursively finalize its value to eliminate intermediate ReferenceValues, ensuring that the original type of om.any is preserved. This patch fixes an issue where, if a ReferenceValue remains a ReferenceValue after being stripped, the finalization process should continue but previously did not (e.g., a list of om.any that is an object returns another list of om.any that is also an object). Signed-off-by: unlsycn <unlsycn@unlsycn.com>
Signed-off-by: unlsycn <unlsycn@unlsycn.com>
See llvm/circt#8325 Signed-off-by: unlsycn <unlsycn@unlsycn.com>
|
Hi! Can this PR be landed? |
|
Yes, I think so. I briefly mentioned it to @uenoku , but we can merge without his approval. |
Further tests will be added later.Done.