Rob Bradford

I'm not sure porting to run on Windows (or any non-Linux host OS) is really viable. We are really taking deep advantage of functionality like epoll, EventFds, as well as...

Looks like the 20.04 Azure LTS kernel was updated from 5.13 to 5.15 (the `prepare_vdpa.sh` script builds a module so we can see the version difference across two builds.)

> > @rbradford what can we do about this in the short term? Seems like we're gonna have to disable `test_vfio` for a while :-/ Maybe we can rely on...

> Then perhaps the best solution right now is to move the vfio tests to bare metal and wait for the regression to be fixed and propagated to Ubuntu. Yeh,...

I identified the following kernel commit as one that impacts on the boot time (I see time to start init from 60ms to 100ms) ``` commit 71020a3c0dff4a00d96922a4a95a067f524a7dcb Author: Thomas Gleixner...

So in the dev-dependencies/pnet case I can see two paths forward to lowering the MSRV: 1. (My preference) remove the pnet dependency 2. Don't test the buildability of tests with...

> Most the versions are updated by bot blindly without considering a big feature. It is good for development but when it comes to production it's a burden. Lots of...

I did some investigation on whether the pnet issue was transient. Unfortunately the newly added dependency on `serde_with` also doesn't compile with 1.56. 1.60 added support optional feature dependencies so...

> > I did some investigation on whether the pnet issue was transient. Unfortunately the newly added dependency on `serde_with` also doesn't compile with 1.56. 1.60 added support optional feature...

> > a.) A dependency requires it, > > We can be a bit more conservative around dependencies. Only real dev dependency updates should call for MSRV updates if they...