Yep, will do. The test does catch it -- it's just not being executed. I think I have a reasonable start on re-enabling surefire + Java tests for SPARK-4159.
On Tue, Dec 9, 2014 at 10:30 PM, Aaron Davidson <[email protected]> wrote: > Oops, that does look like a bug. Strange that the BlockTransferMessageSuite > did not catch this. "+1" sounds like the right solution, would you be able > to submit a PR? > > On Tue, Dec 9, 2014 at 1:53 PM, Sean Owen <[email protected]> wrote: >> >> >> https://github.com/apache/spark/blob/master/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/BlockTransferMessage.java#L70 >> >> public byte[] toByteArray() { >> ByteBuf buf = Unpooled.buffer(encodedLength()); >> buf.writeByte(type().id); >> encode(buf); >> assert buf.writableBytes() == 0 : "Writable bytes remain: " + >> buf.writableBytes(); >> return buf.array(); >> } >> >> Running the Java tests at last might have turned up a little bug here, >> but wanted to check. This makes a buffer to hold enough bytes to >> encode the message. But it writes 1 byte, plus the message. This makes >> the buffer expand, and then does have nonzero capacity afterwards, so >> the assert fails. >> >> So just needs a "+ 1" in the size? > > --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
