mio regressed using a debugger with ipc-channel

[servo]

mio regressed using a debugger with ipc-channel

[jdm]

cc @danlrobertson

[dlrobertson]

You mention mio in the issue, but mio was not used in the commit mentioned. Did you use git bisect to determine that 357abb9 was at fault? If it was the introduction of mio that is causing this, the at fault commit would be f6cefc5.

[jdm]

f6cefc5 makes the panics appear, and be88285 makes them disappear.

[jdm]

(those are the two most recent commits according to git log --topo-order)

[jdm]

My original comment could have been written more clearly; I did not mean to indicate that 357abb9 was the commit at fault.

[dlrobertson]

Thanks that's super helpful! I'll look into it ASAP.

[jdm]

Did you figure anything out, @danlrobertson?

[dlrobertson]

I did some work on this last night, and after some digging around tokio-rs/mio#494 and https://github.com/servo/ipc-channel/blob/master/src/platform/unix/mod.rs#L475-L478 look like likely culprits. Tonight I'm going to try ignoring EINTR.

[nox]

Can we unstuck this to progress on the servo bump?

[dlrobertson]

If you need it ASAP (today) I can pull the tests out and submit a separate PR for that... Otherwise I've got to do some work on that to make it work on Mac. I'd rather keep it as one PR but I totally understand if you need it ASAP

[nox]

Beginning of next week is fine too.