Fixed bug of writing vod playlist

[hori-ryota]

Bugfix of Decoding MediaPlaylist, winsize = 0, Encode has no segments. · Issue #22 · grafov/m3u8

[bradleyfalzon]

If the behaviour here is to support VoD where winsize == 0, perhaps it might be better to write as:

    for i := uint(0); p.Closed || i < p.winsize && count > 0; count-- {

Then a live playlist can still support winsize of 0, but perhaps if the content is VoD, winsize should've been count?

I don't also fully understand (haven't looked yet), why this would get into this state, seems like if it is VoD - winsize should just match count.

[hori-ryota]

p.Closed || i < p.winsize && count > 0 means (p.Closed || i < p.winsize) && count > 0 ?

lgtm, but I feel like having the demand that wants to show everything in live.

[bradleyfalzon]

Sorry, you're right, it should have be written as:

    for i := uint(0); (p.Closed || i < p.winsize && count > 0); count-- {

The original proposal (i < p.winsize || p.winsize == 0) && count > 0 would ignore the winsize when it is zero, but didn't express the fact we're doing this because VoD has a winsize of zero. So I proposed we switch this to (i < p.winsize || p.Closed) && count > 0 which expresses the previous statement a little more clearly. But this would then ignore the winsize of any VoD content, which almost certainly isn't what anyone wants (it's OK for the moment as the winsize cannot be changed, but this behaviour may change in future versions).

So we need to check why the winsize is getting to 0 for VoD content in the first place, and perhaps have it set to count instead, as I believe that's where the root cause is (but I haven't looked closely at this either, seeing an example would help).

[bradleyfalzon]

Actually, regardless of where the underlying issue it, it appears the docs require winsize to be 0 for VoD:

m3u8/structure.go

Line 105 in 11af315

winsize uint // max number of segments displayed in an encoded playlist; need set to zero for VOD playlists

So I think your original proposal is likely the best option for backwards compatibility.

[bradleyfalzon]

In the original issue #22, what's wrong with the originally submitted PR? What makes this PR different?

[hori-ryota]

originally submitted PR write chunk more 1 than the winsize.

[hori-ryota]

Would you merge this?

[bradleyfalzon]

Hi @hori-ryota

Sorry for the delay in this, can you let me know exactly how to reproduce (ideally via an additional test) and I'll verify and merge.

[hori-ryota]

@bradleyfalzon sorry for delay. I added test. how about this?

[grafov]

Sorry for delay too. I checked the code and test and think the bugfix is ok. Thank you @hori-ryota!