You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
PR #3960 added a call to instr_set_label_callback() to set it to NULL
from instr_clone(), but if the callback is non-NULL an assert fires in
that case. This only normally happens with emulation labels that turn
into traces, which happens to not occur in our very few tests of
emulation labels (#3173 covers adding more tests).
We fix that by adding instr_clear_label_callback() here and using that
from instr_clone(), since these callbacks are a little different from
other values and it feels best to not clear them using the set
routine.
A test is added by putting a loop around a scatter-gather expansion,
triggering trace creation. I confirmed that the assert does fire
without this fix with the loop in place.
Fixes#3962
PR #3960 added a call to instr_set_label_callback() to set it to NULL
from instr_clone(), but if the callback is non-NULL an assert fires in
that case. This only normally happens with emulation labels that turn
into traces, which happens to not occur in our very few tests of
emulation labels (#3173 covers adding more tests).
We fix that by adding instr_clear_label_callback() here and using that
from instr_clone(), since these callbacks are a little different from
other values and it feels best to not clear them using the set
routine.
A test is added by putting a loop around a scatter-gather expansion,
triggering trace creation. I confirmed that the assert does fire
without this fix with the loop in place.
Fixes#3962
The implementation of an emulation API requires better test coverage:
#3112
#3104
#3168
The text was updated successfully, but these errors were encountered: