{Source,Buffer}.indexOf(ByteString, Int) should return start index for empty ByteString

[fzhinkin]

For consistency with CharSequence.indexOf (which out functions are mimicking), Source.indexOf(BS, Int) and Buffer.indexOf(BS, Int) should return a start index (0, by default) when byte string to search is empty.

Currently, both functions return 0 (disregard startIndex value).

[fzhinkin]

It is not clear though, how (s as Source).indexOf(ByteString(), 100500) show work.

If it returns 100500 without fetching data, the index may lay outside of actual source's data capacity.
Perhaps, we should still fetch the data?

[JakeWharton]

I don't think you can get away with returning the start index without first requiring at least that much data to be loaded.

The CharSequence one doesn't quite return the start index if the search string is empty and there's not enough data. Doing "short".indexOf("", 10) will return 5 and not 10.

I don't love the idea of returning whatever arbitrary maximum index is available from the Source/Buffer because it can be more non-determinstic as opposed to a fully-static String, so I think the only choice might be to require the data.

[qwwdfsad]

On a side note -- could you please mention fuzzer in issues found by fuzzer? (so the final bounty list would be easier to compile)

[fzhinkin]

After all, I convinced myself that indexOf(empty byte string) should behave the same way CharSequence.indexOf("").
That means, Buffer|Source.indexOf(ByteString, someIndex) == max(0, min(someIndex, buffer.size)).

For Source, we'll try to fetch max(0, someIndex) bytes first.