William Ashley

Related: https://github.com/golang/go/issues/29948 (not a solution, just the same error and discussion around it)

My employer (Amazon) sent an email to [email protected] on Tuesday, Dec 13 to get me added.

My employer (Amazon) sent an email to [email protected] on Tuesday, Dec 13 to get me added.

My employer (Amazon) sent an email to [email protected] on Tuesday, Dec 13 to get me added.

I've run into this to, on an arm linux machine where SAM tries to use an amd64 image for a ```sam build --use-container```. The problem is with ```docker-py```'s [```Container.attach```](https://github.com/docker/docker-py/blob/main/docker/api/container.py#L15), which...

I stripped down that counters.cc file as much as possible to the simplest reproducer. This is aarch64-only (github won't allow attaching a .cc file, so it's renamed to .txt): [minimal-aarch64-counters.txt](https://github.com/rr-debugger/rr/files/13074644/minimal-aarch64-counters.txt)...

I made the following [post](https://lore.kernel.org/linux-perf-users/[email protected]/) to the ```linux-perf-users``` list describing what I've found: ``` armv8pmu: Pending overflow interrupt is discarded when perf event is disabled An issue [1] was opened...

How hard of a dependency does rr have on never missing an overflow signal handler here? Does that assertion failure need to be fatal?

How do you accurately inject these events? Even without missed signals here, these signals' timing isn't exact.

Mark's latest post https://lore.kernel.org/linux-perf-users/ZW9ooVL32WB8U4YB@FVFF77S0Q05N/ has prototype changes to address this, but he also warns it might be too big to be back-ported to older kernels.