fix(HLS): Skip EXT-X-PRELOAD-HINT without full byterange info

[avelad]

Fixes #5089

[avelad]

We need investigate another solutions, for example https://www.akamai.com/blog/performance/-using-ll-hls-with-byte-range-addressing-to-achieve-interoperabi

[github-actions]

Incremental code coverage: 100.00%

[joeyparrish]

If you want to go straight to the recommendation of https://www.akamai.com/blog/performance/-using-ll-hls-with-byte-range-addressing-to-achieve-interoperabi, do this instead in createSegmentReference_:

          // A preload hinted partial segment may have byterange length info.
          const pByterangeLength = item.getAttributeValue('BYTERANGE-LENGTH');
          if (pByterangeLength) {
            pEndByte = pStartByte + Number(pByterangeLength) - 1;
          }
          // *** ADD THIS ***
          else if (pStartByte) {
            // If we have a non-zero start byte, but no end byte, follow the recommendation of
            // https://tinyurl.com/hls-open-byte-range and set the end byte explicitly to a
            // large integer.
            pEndByte = Number.MAX_SAFE_INTEGER;
          }

I haven't tested that against the content in #5089. If it doesn't work, we can just ignore the HINT segment instead.

[avelad]

Yes, I have tried and thought about that, but the problem is that it overlaps with the following partial segments and very rare cases occur. That's why I created this solution.

[avelad]

I think in the future we need to rethink the way we work with byterange, but this is a easy-fix in the meantime.

[joeyparrish]

How does MAX_SAFE_INTEGER on preload hint interact poorly with partial segments?

[avelad]

For a certain segment it has 2 partial segments and 1 preload. In the next update you have 3 partial segments and 1 preload. The preload from before the update has now been converted to a partial segment with all the byterange info, but the new preload doesn't have the info. If we call the first one with the maximum range, we are actually obtaining the corresponding partial and all the following ones.

[joeyparrish]

Ah, I see. It seems that Will Law's article suggests getting and using that continuation on purpose, but Shaka Player isn't built for it at the moment.

[joeyparrish]

Please add a comment here that BYTERANGE-START without BYTERANGE-LENGTH is being ignored.

[avelad]

Done!