On 10/11/2015 21:57, Mark Thomas wrote: > On 10/11/2015 21:54, Rémy Maucherat wrote: >> 2015-11-10 22:34 GMT+01:00 Mark Thomas <ma...@apache.org>: >> >>> Just a quick update on the issues Violeta reported with 9.0.x, HTTP/2 >>> and Windows. >>> >>> I can repeat the Chrome + WebSocket == failure result >>> >> >> What is the expected test result ? If using ALPN, chrome connects using >> HTTP/2, which then has no way to upgrade to something else, including >> websockets. Effectively, this means that websockets applications won't >> work, but this looks normal to me (and it fails also on Linux). > > Ah. That would explain it. We'd need to put WebSockets on a connector > without ALPN support then.
Since I had everything set up, I took a look at the network traffic and all I was seeing was a single request (over h2) for drawboard.xhtml. I did some digging and found the Chrome was reporting errors in that page because no content-type was set. I've fixed that header (and a couple of others) and now the WebSocket examples are working over h2 in Chrome. >>> For ( Chrome | FF ) + stockticker, I see a long pause before any content >>> appears, then I get about 30 lines, then lines start appearing >>> individually as expected. >>> >> >> I didn't notice any issue on stockticker however (also on Linux). > > Yeah, it works on Linux. Not sure what is going on. May be a Windows > network issue. I'll need to fire up Wireshark to check exactly what is > happening when. > > Thanks for the additional info. I've confirmed with Wireshark that the stock prices are sent by Tomcat and I can't see anything different compared to HTTP/1.1. It looks like some sort of buffering issue but I'm not sure where the buffering is taking place. Everything works as expected when I use curl so I suspect a browser buffering issue. Mark --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
