Tim King
I think this is a flake. Unassigning myself for now. I'll take a deeper look if it crops up again.
I do not yet have a good understanding of how these suggestions interact with https://go.dev/ref/spec#Handling_panics . Specifically: > recover was not called directly by a deferred function. How are we...
Here is a crack at handling `defers` and `panics`. The key idea is to record the deferred functions and arguments and to re-raise the panic value if panicking. Need a...
> The problem with deferring in the for-loop after x.Do is that means none of the simulated defer calls will get hit if there's a panic in x.Do or the...
Here is a variant of the previous that uses a defer before x.Do, and within this it defers the _temp_defers and then recovers and re-raises a panic if `recover() !=...
Seems like a great idea to me overall. Minor comments about implementation details. > Whereas the loopclosure check for go, defer, and errgroup.Group.Go only consider invocations in the last statement...
> In the current implementation, we only consider t.Run statements that are at the at the top level of the loop body, and therefore wouldn't check the bodies of those...
@DmitriyMV All 3 of those examples are cyclic. These would be disallowed by the new rules.
One [maybe?] advantage of allowing cyclic interfaces to be anonymous is that they do not require a definition of the interface to be imported in order to be implemented. This...
If the test for that builds knetfile.c is too instrusive/hackish, it can be dropped.