- 10 Feb, 2016 1 commit
-
-
Tatsuhiro Tsujikawa authored
-
- 04 Feb, 2016 4 commits
-
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
- 25 Jan, 2016 3 commits
-
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
- 23 Jan, 2016 2 commits
-
-
Tatsuhiro Tsujikawa authored
This is more accurate, and there is symmetric relation between last_sent_stream_id and last_recv_stream_id, which is bettern fit in my sense.
-
Tatsuhiro Tsujikawa authored
-
- 22 Jan, 2016 4 commits
-
-
Tatsuhiro Tsujikawa authored
Previously, we did terminate session, but it did not actually cancel streams. Now we just cancel streams in this case.
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
- 21 Jan, 2016 9 commits
-
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
By default, as RFC 7540 calls for, pushed stream depends on its associated (parent) stream. There are some situations that this is sub-optimal. For example, if associated stream is HTML, and server is configured to push css and javascript files which are in critical rendering path. Then the default priority scheme is sub-optimal, since browser typically blocks rendering while waiting for critical resources. In this case, it is better to at least give pushed stream the same priority of associated stream, and interleave these streams. This change gives pushed stream the same priority of associated stream if pushed stream has content-type "application/javascript" or "text/css". The pushed stream now depends on the stream which associated stream depends on. We use the same weight of associated stream.
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
This reverts commit e278893b. SSL_CONF_CTX functions are not working correctly with OpenSSL 1.0.2. It requires 1.1.0 to make it work.
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
This is required to avoid creation of temporary ImmutableString like so: std::string x; ImmutableString y = ...; StringRef ref = !x.empty() ? x : y; First, temporary ImmutableString is created with x since ImmutableString has constructor to accept std::string. After StringRef gets this, the temporary ImmutableString is destroyed, and ref has dangling pointer.
-
Tatsuhiro Tsujikawa authored
-
- 20 Jan, 2016 1 commit
-
-
Tatsuhiro Tsujikawa authored
This might help throughput, but it interfere stream priority. The throughput issue is generally caused by the small buffer size to store response body, which was 16K. We increased it to 128K to compensate this change.
-
- 19 Jan, 2016 13 commits
-
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
Because of zero initialization, these assignments are unnecessary.
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
- 18 Jan, 2016 3 commits
-
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-
Tatsuhiro Tsujikawa authored
-