fix(bitswap/client): dont set nil for DontHaveTimeoutConfig

[Wondertan]

I thought I was going nuts... That's not even funny.

[guillaumemichel]

Thanks @Wondertan!

[guillaumemichel]

Actually, since bs.dontHaveTimeoutConfig is only used to define peerQueueFactory, setting it to nil should have no effect on the closure previously defined. This looks like a memory optimization.

Any opinions @gammazero?

[Wondertan]

@guillaumemichel, closure takes value from Client struct tho, so this nil sets the config entirely to zero even if its set.

The DontHaveTimeout component then makes a default value silently if nil is given, so in the end DontHaveTimeout component works but with default config even when user passes a modified config.

This is the reason why I was going "nuts". I spent too much time debugging the config and my understanding of Bitswap code there, while this was an issue...

[gammazero]

Actually, since bs.dontHaveTimeoutConfig is only used to define peerQueueFactory, setting it to nil should have no effect on the closure previously defined. This looks like a memory optimization.

Any opinions @gammazero?

The values inside in the function are not evaluated until the function is run, so when peerQueueFactory is called, bs.perPeerSendDelay will appear to be nil. This is nothing more than an attempt to dispose of memory that is not needed, but it affects what it inside the closure.

[Wondertan]

should we make it skipchangelog?

[gammazero]

should we make it skipchangelog?

I think it would be good make a note that we no longer ignore DontHaveTimeoutConfig. If there is a change in functionality or in the way something is used, then it should be noted in change log. I think this qualifies.