Stream: t-compiler

Topic: spawn_work turns FatalErrors into ICEs ? #66530


pnkfelix (Dec 20 2019 at 14:55, on Zulip):

Hey @mw I had a question for you about our spawn_work architecture

pnkfelix (Dec 20 2019 at 14:56, on Zulip):

the code here is written effectively saying "if a worker thread gets here, it panicked; so we will panic too."

pnkfelix (Dec 20 2019 at 14:57, on Zulip):

but that result can come up when the worker thread hits any FatalError, not just panics or ICEs.

pnkfelix (Dec 20 2019 at 14:58, on Zulip):

(as noted here)

pnkfelix (Dec 20 2019 at 15:00, on Zulip):

is there any reason I should not attempt to turn a FatalError in the worker thread into a corresponding FatalError on the conductor?

pnkfelix (Dec 20 2019 at 15:00, on Zulip):

(so that we get a clean shutdown instead of an ICE from the conductor?)

mw (Dec 20 2019 at 15:02, on Zulip):

Changing this sounds good to me

pnkfelix (Dec 20 2019 at 15:05, on Zulip):

next Q: is that a Major Change? :wink:

mw (Dec 20 2019 at 15:06, on Zulip):

:)

mw (Dec 20 2019 at 15:07, on Zulip):

no, bug fix

Last update: May 26 2020 at 10:25UTC