[irgen] Add support to IRGenDebugInfo for emitting an llvm.dbg.addr instead of llvm.dbg.declare. #41379
Conversation
|
@swift-ci smoke test |
|
The macOS hit a known failure on the smoke test bots. I am going to wait for #41378 to land and then I am going to add a bit more into this PR including a full end to end test that shows that we get the correct debug info. |
…nstead of llvm.dbg.declare. By default, we still use llvm.dbg.declare, so this is an NFC commit. But this sets us up for using this in the future.
gottesmm
force-pushed
the
pr-aa7d71cdb8ea8518b4d7e9b78a9f99d91639c1c0
branch
from
498d39d to
2597c56
Compare
|
@swift-ci test |
|
The extra commit is 2597c56. The patch makes sure that the debug_value undef that we insert uses the exact location as the original value that we are moving. Otherwise, LLVM doesn't seem to handle this correctly. I also included an end to end test that checks that we handle debug info for move of let/var for copyable/address only correctly. In the process I discovered that the let object works as expected but the address only let/copyable & address only var are not working for some reason. I am looking into it now, but I decided to upstream the test with the file check lines for the let case but not the other ones. I am going to fix that in a subsequent PR. |
…riginal DbgVarCarryingInst we found. LLVM wants the debug_value undef to be /exactly/ the same as the original debug_value including the loc. If we don't match, LLVM won't invalidate the value. Also now that things actually work (at least for copyable lets), lets add some basic tests that debug info works as expected. NOTE: With this patch it seems that address only move is still broken. I am looking into it and will fix it in a forthcoming PR.
gottesmm
force-pushed
the
pr-aa7d71cdb8ea8518b4d7e9b78a9f99d91639c1c0
branch
from
2597c56 to
504fe52
Compare
|
@swift-ci test |
By default, we still use llvm.dbg.declare, so this is an NFC commit. But this
sets us up for using this in the future.