Improved socket implementation to recycle buffer in Socket.readXX ope… #809
Conversation
|
@mpilquist this complains about the binary compatibility, but hence the change is not in api, I still believe its ok to publish it in 0.9.3 any thoughts? |
|
@pchlupacek Yeah that's fine -- can you add an exclusion to |
| getBufferOf(max) flatMap { buff => | ||
| readChunk(buff, timeout.map(_.toMillis).getOrElse(0l)) flatMap { | ||
| case (read, _) => | ||
| if (read < 0) F.pure(None) |
There was a problem hiding this comment.
The readSemaphore isn't incremented here -- is that okay?
There was a problem hiding this comment.
thanks for spotting this
There was a problem hiding this comment.
but actually, hence this signals eof, i think we won't hit it,neverthelss I gonna fix it for corectness
| if (readBytes < 0 || buff.remaining() <= 0) F.pure(Some(Chunk.bytes(buff.array(), 0, buff.position()))) | ||
| else go(buff,(timeoutMs - took) max 0) | ||
| def read0(max:Int, timeout:Option[FiniteDuration]):F[Option[Chunk[Byte]]] = { | ||
| readSemaphore.decrement >> |
There was a problem hiding this comment.
Do we need any error handling here? E.g., if one of these tasks throws, we still want to increment the semaphore, right?
There was a problem hiding this comment.
Well I think if you fail the read, there is no way to recover it anyhow, so my thought was just to let it be. Do you see there is a way that read will fail, and there is chance to recover it with next read?
|
@mpilquist made modifications as you have suggested. Not sure about this mimaBinaryIssueFilters where they are? |
|
@pchlupacek I took care of the mima issue: b02fd92 |
|
@mpilquist thanks |
…rations