Skip to content

inference: fix missing LimitedAccuracy markers #55362

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Aug 5, 2024
Merged

inference: fix missing LimitedAccuracy markers #55362

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 15 additions & 5 deletions base/compiler/abstractinterpretation.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -698,13 +698,23 @@ function abstract_call_method(interp::AbstractInterpreter,
end
add_remark!(interp, sv, washardlimit ? RECURSION_MSG_HARDLIMIT : RECURSION_MSG)
# TODO (#48913) implement a proper recursion handling for irinterp:
# This works just because currently the `:terminate` condition guarantees that
# irinterp doesn't fail into unresolved cycles, but it's not a good solution.
# This works just because currently the `:terminate` condition usually means this is unreachable here
# for irinterp because there are not unresolved cycles, but it's not a good solution.
# We should revisit this once we have a better story for handling cycles in irinterp.
if isa(topmost, InferenceState)
if isa(sv, InferenceState)
# since the hardlimit is against the edge to the parent frame,
# we should try to poison the whole edge, not just the topmost frame
parentframe = frame_parent(topmost)
if isa(sv, InferenceState) && isa(parentframe, InferenceState)
poison_callstack!(sv, parentframe === nothing ? topmost : parentframe)
while !isa(parentframe, InferenceState)
# attempt to find a parent frame that can handle this LimitedAccuracy result correctly
# so we don't try to cache this incomplete intermediate result
parentframe === nothing && break
parentframe = frame_parent(parentframe)
end
if isa(parentframe, InferenceState)
poison_callstack!(sv, parentframe)
elseif isa(topmost, InferenceState)
poison_callstack!(sv, topmost)
end
end
# n.b. this heuristic depends on the non-local state, so we must record the limit later
Expand Down