Skip to content

Make pseq robust by using seq#

Matthew Craven requested to merge wip/T22935 into master

Fixes #23699, fixes #23233. Note [seq# magic] is expanded, fixing #22935.

The two tests for T15226 do not really test issue #15226 (closed); instead they check that we can unbox through seq# in some conditions. The way we previously achieved that for these two test cases is generally bogus; see the new wrinkle S6 in Note [seq# magic]. Safely unboxing through seq# is probably possible under the right conditions but deserves a new primop and is left as future work.

Merge request reports