fix(safari): initial seek time not taken into account with fromLastPosition

[Florent-Bouisset]

fix a bug on Safari where a live stream will play at the time=0 rather than next to the live edge when using startAt:fromLastPosition. This appears on some contents that have different protection Key for each of their track.

[peaBerberian]

For the JSDoc:

1. Here it's number, not string
2. there's no space in the type annotation
3. Also, there may be a dash between the variable name and its description, that I prefer to add for readability personally, as you wish
4. In the param documentation, for undefined, I don't think just repeating the current implementation has value in this case, it just repeats what the current code does.

I'm wondering about (4), if the undefined seeking time notion leading to a readyState of 2 is not too specific for the general compat files.
The source undefined is returned by getInitialTime when it cannot determine the initial time, but I feel that making compat return 2 here so the RxPlayer logic may check later with getInitialTime again is awkward: compat has no idea what the initial-seek logic is.

To me compat should just say: this things doesn't work on the current device, or this value shoul be used on the current device. It's kind of what the function name (getMinimumReadyStateForSeeking) is suggesting but checking for the first getInitialTime's return value in it is kind of out-of-place I feel.

For implementation-related details such as "we will re-call getInitialTime later", it should be constrained to the location where getInitialTime is called IMO.

[Florent-Bouisset]

Yes will update the formating / type error, the formatter does not format JSDoc right ?

I agree with what you are saying with the concept of initial time does not belong to the compat folder.

Do you think this is better ?

// compat
getMinimumReadyStateForSeeking() {
   return isSafari ? 2 : 1;
}

// initial_seek_and_play
const initiallySeekedTime = typeof startTime === "number" ? startTime : startTime();
const minReadyStateForSeeking = getMinimumReadyStateForSeeking();

The difference is that it will always seek at readyState 2 depending or not if the seek time is known or undefined

[peaBerberian]

formatter does not format JSDoc right ?

I don't think so

Do you think this is better ?

For me the issue is not that there's a minimum ready state for seeking, it's 1 for both in absolute, it's more that 1 does not reliably means a known duration on safari. I think the compat util should reflect that.

[peaBerberian]

I generally like to add a namespace: at the beginning of logs to simplify treatments (e.g. by filering/excluding some "family" of logs) and indicate quickly where the logic runs.
Here we could prepend it with Init:

Though this is not done with the other warn here, I don't know why. It seems to however be done with other calls on that file.

[Florent-Bouisset]

Yep I basically copy/paste the existing log
I'm adding the prefix for the one missing

[peaBerberian]

You have https://developer.mozilla.org/en-US/docs/Web/API/HTMLMediaElement/readyState#htmlmediaelement.have_current_data that allow to remove the need to have a comment here