Dan Kortschak

I'd like to see what impact this has on time spent in `runtime.scanobject`. I've done some experiments with a family of `type T1 struct { next *T1; pad [n]byte }`...

This is the output of the benchmarking that I did. Apart from very large cases (padding with 2^16 bytes) there is no difference in behaviours between front loaded pointers and...

I agree that if the changes made a difference that it would be significant given the proportion of the time spent in `scanobject`, but the change in ordering doesn't seem...

On the basis of experiments that I did to see the performance impact of pointer byte count (essentially none) I'm having second thought about this.

I think this is now done.

The linter failure looks like it is due to errorlint incorrectly handling cases where it expects a fmt verb but there is none. This should not be true, but in...

It's worth noting that this is a general failure though I suppose not many people use indexed verbs. I have filed an issue against polyfloyd/go-errorlint which is now fixed there...

The logic that is causing this is as follows: 1. A security sensitive command is invoked with secrets on the command line. 2. The command is aware of the sensitivity...

After discussion with others on the team, I'm going to close this as working as intended; the information is not available because a security-aware application has scrubbed the process's `argv`.

I have the same concerns about the breaking change. The impetus for this is that the ECS specifies the format used in this change.