test_gap_fetch_handle_lifecycle has failed on main since #11682 (b67ad0005, "honest handles: EventTarget family are real objects; built-in constructor statics read the constructor").
Bisect on perrymaster (release builds, PERRY_SKIP_BUILD=1 PERRY_NO_AUTO_OPTIMIZE=1 ./run_parity_tests.sh --filter test_gap_fetch_handle_lifecycle, Node 26.5.1, 2 runs each):
It is deterministic. It also fails 4/4 on main 23d563404.
The diff against Node is on line 4:
< request body response body subclass body
> request body response body {}
class RetainedResponse extends Response {}: the subclass instance's body now reads back as {} instead of "subclass body". This looks like #11682's Response/EventTarget honest-handle change losing a subclass wrapper's handle. #11682's own PR CI showed gap-suite (1) and gap-suite (5) red.
Found while triaging #11676. The same failure appears on #11694 and every PR tested against current main.
test_gap_fetch_handle_lifecyclehas failed on main since #11682 (b67ad0005, "honest handles: EventTarget family are real objects; built-in constructor statics read the constructor").Bisect on perrymaster (release builds,
PERRY_SKIP_BUILD=1 PERRY_NO_AUTO_OPTIMIZE=1 ./run_parity_tests.sh --filter test_gap_fetch_handle_lifecycle, Node 26.5.1, 2 runs each):7fa094cb4(parent): passes 2/2.b67ad0005(honest handles: EventTarget family are real objects; built-in constructor statics read the constructor #11682): parity_fail 2/2.It is deterministic. It also fails 4/4 on main
23d563404.The diff against Node is on line 4:
class RetainedResponse extends Response {}: the subclass instance's body now reads back as{}instead of"subclass body". This looks like #11682's Response/EventTarget honest-handle change losing a subclass wrapper's handle. #11682's own PR CI showedgap-suite (1)andgap-suite (5)red.Found while triaging #11676. The same failure appears on #11694 and every PR tested against current main.