On Mon, Jul 01, 2019 at 07:17:39PM -0400, John Snow wrote: > > > On 7/1/19 5:27 PM, Eduardo Habkost wrote: > > On Thu, Jun 27, 2019 at 05:32:17PM -0400, John Snow wrote: > >> I didn't actually mean to retain the RFC tag, but oh well. > > > > I'm queueing patch 1-2 for QEMU 4.1. > > > > Patch 3 will require more careful review. If you can demonstrate > > it fixes a problem in some use cases, we can call it a bug fix > > and include it after soft freeze. ;) > > > > Right. > > Mostly, it's that it ignores things like ctrl^c -- it ignores ANY > failure. I don't think that it should.
Agreed. > > In cases where QEMU crashes, the caller will now get an exception -- but > the cleanup will still faithfully execute. There are some exceptions we > may genuinely wish to suppress here, but it's hard to enumerate which > ones without going too broadly. > > It's fine to exclude this for now, but I think it's still something we > want eventually. We just need to ensure all callers of shutdown() are ready for this change. I know that scripts/device-crash-test expects shutdown() to not raise an exception if QEMU crashed, inside checkOneTestCase(). Maybe that's the only caller that needs to change. -- Eduardo