Cosmin Apreutesei
2014-07-18 00:37:12 UTC
Hi,
Sorry in advance for the "dissatisfied customer" tone (this doesn't
happen as often as the tone implies).
Here's my situation:
My callback-ridden code sometimes panics with "bad callback", and it
makes it hard to trust that it won't break on a user's machine because
of a slightly different call order that I can't reproduce easily (and
unit-test for) due to the async nature of the issue. Also, for panics
that happen rarely, it's hard to really be sure that I even applied
jit.off() to the right function. It kinda makes me paranoid.
I would like to find a way to make these panics more deterministic,
eg. a compile option to disable the automatic jit guards would maybe
make them trigger more rapidly and/or often?
Mike, any tips?
What do people do when confronted with this?
Thanks.
Sorry in advance for the "dissatisfied customer" tone (this doesn't
happen as often as the tone implies).
Here's my situation:
My callback-ridden code sometimes panics with "bad callback", and it
makes it hard to trust that it won't break on a user's machine because
of a slightly different call order that I can't reproduce easily (and
unit-test for) due to the async nature of the issue. Also, for panics
that happen rarely, it's hard to really be sure that I even applied
jit.off() to the right function. It kinda makes me paranoid.
I would like to find a way to make these panics more deterministic,
eg. a compile option to disable the automatic jit guards would maybe
make them trigger more rapidly and/or often?
Mike, any tips?
What do people do when confronted with this?
Thanks.