gh-112355: fix calculation of jump target of ENTER_EXECUTOR in dis #112377
Conversation
| @@ -373,6 +374,8 @@ def _get_argval_argrepr(op, arg, offset, co_consts, names, varname_from_oparg, | |||
| argval = offset + 2 + signed_arg*2 | |||
| caches = _get_cache_size(_all_opname[deop]) | |||
| argval += 2 * caches | |||
| if deop == ENTER_EXECUTOR: | |||
| argval += 2 | |||
There was a problem hiding this comment.
It would be nice if we didn't need to special case this here, but that would require a change in the oparg (and implementation) of ENTER_EXECUTOR, and I figured this PR should just fix the bug in dis.
There was a problem hiding this comment.
Oh, this must be the cache of the underlying JUMP_BACKWARD. I think there currently is no way to get that. Not even via _testinternalcapi.
we should file an issue about that, it will get worse once we start putting executors on other ops.
There was a problem hiding this comment.
Created an issue here: #112383
| @@ -373,6 +374,8 @@ def _get_argval_argrepr(op, arg, offset, co_consts, names, varname_from_oparg, | |||
| argval = offset + 2 + signed_arg*2 | |||
| caches = _get_cache_size(_all_opname[deop]) | |||
| argval += 2 * caches | |||
| if deop == ENTER_EXECUTOR: | |||
| argval += 2 | |||
There was a problem hiding this comment.
Oh, this must be the cache of the underlying JUMP_BACKWARD. I think there currently is no way to get that. Not even via _testinternalcapi.
we should file an issue about that, it will get worse once we start putting executors on other ops.
With this
./python.exe -Xuops -m test -v test_dispasses.