Xiangyu Zhu

Same here with chrome 92.0.4515.131. I disabled kwallet so failing to invoke `keyring.get_password` is expected (not sure if the behavior is correct though). I commented out the relevant part and...

Thanks for all the suggestions & comments. Code has been updated.

godoc & others updated. For the benchmark, during development I ran 2 comparison check to see the performance impact: ### default clock One is to compare the default clock against...

@MrAlias Thanks for the suggestion, here's what I have in mind. For a typical offset clock with the current `Clock` interface, we can ensure monotonic start/end times as long as...

I have made some modifications regarding the suggested direction given by @MrAlias . Probably there are still many places that can be improved, though. To summary a few important points...

Although I personally still prefer the previous interface because I think the correctness of endtime (when `Stop` called) should leaves to the custom clock implementation (for default clock it should...

Updated. Here's the current benchmark result. From a quick look there seems to have a constant 30~70 ns/op overhead (maximum ~10% for simple span, maximum ~3% for span with 10...

I added a parameter to clock's `Stopwatch` method (for a standard default clock this basically moves the `IsZero` check from the caller into the clock's `Stopwatch` method). For a custom...

Sorry for the long pause from last update. We still think this feature might be useful in some cases (although to be honest not very useful to most people). I...

Here a configurable "properties" to be keyed is wanted, but my PR only keyed the message by traceID though (intended to be used with tail-based sampling). Maybe that PR can...