Dave Thaler

The BSD sockets API is cross-plat and so a key sockets scenario is cross-plat. Hence this should not be a question, it should simply be a feature request.

ebpf-for-windows project is now also blocked updating beyond 1.0.3 due to this issue.

> @dthaler have you tested with v1.0.4? Yes, we see intermittant failures, such as in https://github.com/microsoft/ebpf-for-windows/runs/5726574449?check_suite_focus=true (whereas https://github.com/microsoft/ebpf-for-windows/runs/5725496349?check_suite_focus=true passed referencing the same commit)

FYI, in the ebpf-for-windows project, I used `dependabot ignore this minor version` to block upgrading the scorecard-action from 1.0.3 until this issue is fixed or 1.1 comes out, since the...

@laurentsimon > We will try to add PR support in the next release. Now that v2.0 is out, is there any progress on fixing this PR?

> @dthaler I have merged #384 by mistake. I can revert if that will be helpful. No need, this one just tests the "broken" functionality, and 384 fixes the code...

Poorna pointed me to https://cmake.org/cmake/help/latest/command/file.html which says: > **Note** > > We do not recommend using GLOB to collect a list of source files from your source tree. If no...

PR #212 fixed most of the issues. The following helpers still remain unchecked after that PR: - tail_call: void * - override_return: struct pt_reg * - perf_event_output: struct pt_reg *...

I believe this can (and should) be done in a generic fashion that would also support things like: * [bpf_spin_lock](https://www.man7.org/linux/man-pages/man7/bpf-helpers.7.html)() and bpf_spin_unlock() helpers must be paired and any other such...

I think the spin_lock case of only-one acquire is easily handled by the runtime, especially given the man page says: > The BPF program can take ONE lock at a...