3d863ed254
nghttp2_hd: Clear reference set with index 0
2013-11-16 17:05:18 +09:00
5696a65c5e
nghttp2_hd: Update doc
2013-11-16 16:33:38 +09:00
94ec4dbe42
nghttpd_hd: Update static table and use 1-based index
...
The specification now says that index to the header table entry is
1-based. Since 0-based index is very handy to access arrays, we
internally uses 0-based index. We just convert it to 1-based when
we emit to the block and convert 1-based to 0-based on decoding.
2013-11-16 16:18:44 +09:00
1f3840f4a2
Update doc
2013-11-15 00:57:15 +09:00
40347487c9
Don't sort headers in library code
...
Remove sorting headers from library code. The application must sort
them if necessary. nghttpx and nghttpd do the sorting of the headers
in stable way if names are equal.
2013-11-13 23:56:02 +09:00
45c3c5b80f
Cleanup nghttp2_session_mem_recv
2013-11-12 10:52:03 +09:00
a39a816a63
Cleanup nghttp2_session_next_data_read
2013-11-12 10:48:29 +09:00
5044b59453
Handle error from nghttp2_session_push_back_deferred_data
2013-11-12 10:44:04 +09:00
415b8ed674
Code cleanup
2013-11-08 01:47:29 +09:00
ac354d3b91
Fix compile error with --enable-maintainer-mode
2013-11-08 01:35:15 +09:00
2ae788eddd
Replace nghttp2_set_option with nghttp2_session_{client,server}_new2
...
nghttp2_session_client_new2 and nghttp2_session_server_new2 take
additional parameters which specifies session options.
nghttp2_set_option is somewhat crumsy because of type checking.
Now we use nghttp2_opt_set, which specifies individual options with
types. We changed the value of nghttp2_opt, so this change will
require re-compile.
2013-11-08 00:12:39 +09:00
d92a161c46
nghttp2_hd_huffman: Code cleanup
2013-11-05 23:34:58 +09:00
0d6faa74b7
nghttp2_hd_huffman: Avoid memset on encoding
2013-11-03 22:31:06 +09:00
2d08d30409
nghttp2_hd: Fix crash
2013-11-03 17:46:10 +09:00
4bc44b0c0b
Implement SETTINGS_ENABLE_PUSH
...
It is not clear that SETTINGS_ENABLE_PUSH = 0 disallows HEADERS
to the reserved streams. For now, we just check the reception
and transmission of PUSH_PROMISE against SETTINGS_ENABLE_PUSH.
2013-11-02 16:53:06 +09:00
29ef3fde8b
Initialize settings in common function
2013-11-02 16:23:56 +09:00
66fce93d98
Update doc
2013-10-31 02:12:48 +09:00
b0f76773e1
nghttp2_hd: Use defalte_hd_table_bufsize for decoder as well
2013-10-31 01:51:16 +09:00
c8a9f8d312
nghttp2_hd: Fix crash on fail
2013-10-31 01:26:37 +09:00
ad64f11a95
Remove code submitting WINDOW_UPDATE with negative delta
...
It is effectively dead code because we should have already submit
WINDOW_UPDATE before this case happens.
2013-10-30 23:43:40 +09:00
2e7bc014e1
Return 0 if recv_window_size is negative; add tests
2013-10-30 01:58:38 +09:00
dfa1194804
Add new API to return effective recv data/win size for connection
...
Using this feature, connection level flow control is now enabled
in nghttpx.
2013-10-30 00:51:01 +09:00
6c23c34d77
Add new API to return effective received data length and local window size
2013-10-30 00:07:35 +09:00
a4cf37306a
Enfoce flow control error strictly
...
Now we have SETTINGS synchronization, flow control error can be
detected strictly. If DATA frame is received with length > 0 and
current received window size is equal to or larger than local
window size (latter happens when we shirnk window size), it is
subject to FLOW_CONTROL_ERROR,
2013-10-29 22:46:30 +09:00
6c99ff12c9
nghttp2_hd: Rename local as deflate
2013-10-29 00:42:08 +09:00
683253e334
Declare HTTP-draft-07/2.0
2013-10-27 23:16:49 +09:00
bfa7dfb37e
Use bit shift for easy reading
2013-10-27 23:09:10 +09:00
34b3833c71
Update doc
2013-10-27 21:27:48 +09:00
67ce60544e
Change maximum frame length to 16383
2013-10-27 21:17:09 +09:00
cfb9857f27
Use FRAME_SIZE_ERROR for invalid payload length
2013-10-27 20:55:44 +09:00
431f78f008
Rename NGHTTP2_ERR_FRAME_TOO_LARGE as NGHTTP2_ERR_FRAME_SIZE_ERROR
2013-10-27 19:35:36 +09:00
5ace898392
Update error code
2013-10-27 19:31:24 +09:00
a46ccdb144
Apply initiated SETTINGS changes on reception of ACK
2013-10-27 19:22:51 +09:00
22df6f4190
nghttp2_hd: Fix bug on local_hd_table_bufsize_max > hd_table_bufsize_max
2013-10-26 23:30:48 +09:00
cbdd44c4ae
nghttp2_hd: Implement local header table size limit for encoder
2013-10-26 18:49:23 +09:00
8f8c841df0
Merge branch 'master' into hpack-exp
2013-10-25 22:51:51 +09:00
896db5b24f
API change: Add flags parameter to all submit_* functions
...
The nghttp2_submit_{request,response} family do not get this change.
2013-10-25 22:50:24 +09:00
abfc00983d
nghttp2_hd: Fix memory leak and bad free
2013-10-24 23:49:37 +09:00
0c3cb104e9
nghttp2_hd: Share static table globally
2013-10-24 22:09:40 +09:00
5a81e03497
nghttp2_hd: Add static table entry to dynamic table on emission
2013-10-24 21:52:02 +09:00
d5ccc89fc4
Update huffman table and static table
...
Use huffman table in https://github.com/grmocg/httpbis-header-compression
2013-10-21 23:56:14 +09:00
9e50ae46d9
nghttp_hd: Use NGHTTP2_HD_SIDE_{REQUEST,RESPONSE} instead of {CLIENT,SERVER}
...
This change conveys better notion about compression context.
2013-10-21 00:44:39 +09:00
8ef134a702
nghttp_hd: Refactor emit_indname_block and emit_newname_block
2013-10-21 00:22:14 +09:00
28aea5c6b7
nghttp_hd: Use huffman encoding only when it is usable
...
We use huffman encoding only when its encoded length is strictly less
than the original length.
As noted in the HPACK draft, the length of name/value is now 7-bit
prefix and its MSB is 1 if the following string is huffman encoded.
2013-10-20 23:59:15 +09:00
a658f1367e
encode_length: Or-ing first byte with prefix mask
...
So that we can preserve leading bits.
2013-10-20 16:28:52 +09:00
4e4fab01b1
Merge branch 'master' into hpack-exp
...
Conflicts:
lib/nghttp2_hd.c
2013-10-18 20:06:36 +09:00
7b87d71121
nghttp2_hd: Fix missing return value handling
2013-10-18 19:48:15 +09:00
0efa6e657f
Fix outbound flow control count
...
We wrongly added the whole payload length even if we sent part of it.
2013-10-18 19:43:59 +09:00
7bd145fd23
Fix out of bound array access in decode_length
2013-10-18 19:30:04 +09:00
f7389ff2e6
Revert 622b05aa31
...
It turns out that 0-bit prefix is wrong, and the author now
clearly stated that the intention is 8-bit prefix for 8+ fields.
2013-10-18 19:28:01 +09:00
Tatsuhiro Tsujikawa