Possible race condition in context propagation about atreugo

[gurich92]

@dmarkhas
Also faced that race condition.

For workaround I used the underlying fasthttp.RequestCtx, since if my understanding is correct - it's main context here, because atreugo.RequestCtx is just a top-level "abstraction"(

atreugo/types.go

Line 387 in 96c313b

// RequestCtx context wrapper of fasthttp.RequestCtx to adds extra funtionality

), but for all the routing, timeouts etc standard fasthttp.Router and so fasthttp.RequestCtx is used.

So your code should work well if it will be like this:

func(ctx *atreugo.RequestCtx) error {

		newCtx, cancel := context.WithTimeout(ctx.RequestCtx, time.Second*1)
		defer cancel()

		return ctx.TextResponse("OK")
}

[savsgio]

Hi @dmarkhas,

Sorry for delayed answer.

I will try to research about this race condition, but in the meanwhile, try to use the underlying ctx.RequestCtx, just like @gurich92 says.

[savsgio]

Hi @dmarkhas and @gurich92,

Sorry for delayed answer, but I've been busy.

I researched about this race condition and could see that fasthttp.RequestCtx (and also atreugo.RequestCtx) are unsafed to use concurrently (https://github.com/valyala/fasthttp/blob/master/server.go#L554).

So, due to the context.WithTimeout() could create a gorutine to check the state. It's better to use context.Backround() instead of the RequestCtx.

Thanks!