agentzh (Yichun Zhang)
|
67f8bbbde3
|
removed the valgrind suppression rules for sendmsg/ngx_channel_write since we already fix these issues via the channel-uninit-params patch: https://github.com/agentzh/ngx_openresty/blob/master/patches/nginx-1.2.3-channel-uninit-params.patch
|
2012-09-17 11:48:04 -07:00 |
|
agentzh (Yichun Zhang)
|
2e50b19433
|
updated valgrind.suppress for valgrind 3.8.0.
|
2012-09-15 21:23:22 -07:00 |
|
agentzh (Yichun Zhang)
|
66e2d98b05
|
docs: bumped version number to 0.6.5 and also documented the trick for doing background async jobs by using ngx.eof() + keepalive_timeout 0.
|
2012-09-15 13:16:22 -07:00 |
|
agentzh (Yichun Zhang)
|
0c14852af3
|
suppressed the warnings reported by valgrind 3.8.0.
|
2012-09-14 14:51:08 -07:00 |
|
agentzh (Yichun Zhang)
|
5b5a02dbed
|
fixed a test case in 068-socket-keepalive.t by properly flushing the redis store first.
|
2012-09-13 15:21:49 -07:00 |
|
agentzh (Yichun Zhang)
|
62a0b69609
|
updated docs to reflect recent changes.
|
2012-09-13 11:46:55 -07:00 |
|
agentzh (Yichun Zhang)
|
b1289817d9
|
added (passing) tests for using parallel subrequests after ngx.eof().
|
2012-09-12 20:45:53 -07:00 |
|
agentzh (Yichun Zhang)
|
e0adfe5ec4
|
bugfix: the UDP and TCP cosockets could not keep track of the current Lua coroutine. now we also explicitly clear the current Lua coroutine context in the main request context to ensure that the I/O scheduler can indeed keep track of the current Lua coroutine itself.
|
2012-09-12 17:06:14 -07:00 |
|
agentzh (Yichun Zhang)
|
e49622d8b6
|
added a (passing) test case for using ngx.flush(true) in a user coroutine.
|
2012-09-12 15:21:27 -07:00 |
|
agentzh (Yichun Zhang)
|
fe8d7be132
|
optimize: we now skip processing in the default write event handler when the write event is not ready.
|
2012-09-12 15:08:05 -07:00 |
|
agentzh (Yichun Zhang)
|
79dee2697a
|
coding style fixes and also fixed a warning in clang.
|
2012-09-12 14:45:39 -07:00 |
|
agentzh (Yichun Zhang)
|
2fbb4ada99
|
refactor: removed the "cur_co" field from ngx_http_lua_ctx_t because it is not necessary at all.
|
2012-09-12 14:43:45 -07:00 |
|
agentzh (Yichun Zhang)
|
644b0b15c9
|
cleaned up the code a bit in ngx_http_lua_wev_handler.
|
2012-09-12 12:04:32 -07:00 |
|
agentzh (Yichun Zhang)
|
240caaa72e
|
refactor: made ngx.flush(true) keep track of the current Lua coroutine automatically.
|
2012-09-11 22:38:37 -07:00 |
|
agentzh (Yichun Zhang)
|
59c5231b8c
|
refactor: renamed the confusing "waiting" field of ngx_http_lua_co_ctx_t to "pending_subreqs".
|
2012-09-11 19:34:53 -07:00 |
|
agentzh (Yichun Zhang)
|
9155f9400d
|
refactor: made ngx.req.read_body() keep track of the current Lua coroutine automatically.
|
2012-09-11 17:55:40 -07:00 |
|
agentzh (Yichun Zhang)
|
6fba5e1bff
|
refactor: made ngx.req.socket() use the existing req_body_reader_co_ctx field in the context struct.
|
2012-09-11 17:32:47 -07:00 |
|
agentzh (Yichun Zhang)
|
12981c9044
|
refactor: made the ngx.req.socket() objects keep track of the current Lua coroutine themselves. also disabled the ngx.req.socket() API in subrequests to prevent potential problems.
|
2012-09-11 17:22:09 -07:00 |
|
agentzh (Yichun Zhang)
|
357d4d7613
|
refactor: made the UDP/datagram cosockets schedule their I/O automatically on the Lua coroutine level.
|
2012-09-11 16:34:32 -07:00 |
|
agentzh (Yichun Zhang)
|
b88aa3eaff
|
refactor: removed the unused "socket_busy" flag from ngx_http_lua_co_ctx_t.
|
2012-09-11 16:14:34 -07:00 |
|
agentzh (Yichun Zhang)
|
9cb0d829d3
|
refactor: made TCP/stream cosockets schedule their I/O automatically on the Lua coroutine level.
|
2012-09-11 16:05:09 -07:00 |
|
agentzh (Yichun Zhang)
|
e2f8b2a6b4
|
enabled the http-lua-info static probes in ngx_http_lua_wev_handler.
|
2012-09-11 12:14:00 -07:00 |
|
agentzh (Yichun Zhang)
|
2f884b83e0
|
refactor: made ngx.location.capture() and ngx.location.capture_multi schedule themselves at the coroutine level (instead of the request level).
|
2012-09-10 19:04:45 -07:00 |
|
agentzh (Yichun Zhang)
|
27dd2bab82
|
minor cleanup in ngx.sleep().
|
2012-09-10 15:44:09 -07:00 |
|
agentzh (Yichun Zhang)
|
7134d428a7
|
refactor: made ngx.sleep() schedule itself at the coroutine level.
|
2012-09-10 15:23:12 -07:00 |
|
agentzh (Yichun Zhang)
|
4abd896cd5
|
fixed a test case in 020-subrequest.t for slow machines.
|
2012-09-10 12:05:49 -07:00 |
|
agentzh (Yichun Zhang)
|
85c658475a
|
made the new subrequest test cases less possible to fail unexpectedly on slow machines.
|
2012-09-10 01:03:44 -07:00 |
|
agentzh (Yichun Zhang)
|
9fe21af577
|
use google.com instead of taobao.com in the test suite.
|
2012-09-10 00:59:14 -07:00 |
|
agentzh (Yichun Zhang)
|
a7d6962c21
|
upgraded version to 0.6.4.
|
2012-09-09 21:30:41 -07:00 |
|
agentzh (Yichun Zhang)
|
e572d87726
|
fixed the test cases for the mockeagain (W) testing mode.
|
2012-09-09 16:01:24 -07:00 |
|
agentzh (Yichun Zhang)
|
73faa581d8
|
added more tests for bad upstream in subrequests.
|
2012-09-09 12:57:35 -07:00 |
|
agentzh (Yichun Zhang)
|
b3255acdae
|
bugfix: reset the subrequest status code when the non-buffered ngx_http_upstream request in the subrequest fails due to timeout errors or premature connection close and etc. this requires the patch for the nginx core to fix an issue in ngx_http_upstream: https://raw.github.com/agentzh/ngx_openresty/master/patches/nginx-1.2.3-nonbuffered-upstream-truncation.patch
|
2012-09-08 22:37:55 -07:00 |
|
agentzh (Yichun Zhang)
|
d06e260181
|
made a test case requires DNS resolving less possible to fail in slow testing modes.
|
2012-09-07 14:32:09 -07:00 |
|
agentzh (Yichun Zhang)
|
c56a98cd80
|
bugfix: we should not always return NGX_OK in our post_subrequest handler even if rc >= 300 because it will prevent response headers from being sent.
|
2012-09-05 22:11:33 -07:00 |
|
agentzh (Yichun Zhang)
|
865cb1981b
|
added a (passing) test for returning NGX_ERROR in the output body filter for the subrequest issued by ngx.location.capture().
|
2012-09-05 21:47:13 -07:00 |
|
agentzh (Yichun Zhang)
|
46a3c3d524
|
updated .gitignore a bit.
|
2012-09-05 17:40:19 -07:00 |
|
agentzh (Yichun Zhang)
|
9728454afd
|
bugfix: the main request might be prematurely terminated if a subrequest issued by ngx.location.capture (or its friends) was finalized with error codes.
|
2012-09-05 17:39:03 -07:00 |
|
agentzh (Yichun Zhang)
|
75c8c30980
|
updated valgrind.suppress for i386.
|
2012-09-05 16:15:40 -07:00 |
|
agentzh (Yichun Zhang)
|
3a90b57408
|
bugfix: we did not compile with nginx cores without the allow_request_body_updating patch. thanks Dirk Feytons.
|
2012-08-30 11:48:21 -07:00 |
|
agentzh (Yichun Zhang)
|
3b6f584be3
|
tweaked a test case to make it less possible to fail in slow testing modes.
|
2012-08-29 20:16:48 -07:00 |
|
agentzh (Yichun Zhang)
|
fd0a8fb182
|
refactor: now ngx_lua issues subrequests without updating r->postponed at all so as to allow the parent requests flush outputs immediately without waiting for its running subrequests. this will pave a way for subrequest integration with the upcoming lightweight thread model.
|
2012-08-29 18:08:24 -07:00 |
|
agentzh (Yichun Zhang)
|
7819b4284a
|
fixed typos in the test titles.
|
2012-08-29 12:29:15 -07:00 |
|
agentzh (Yichun Zhang)
|
ce1bcd53cb
|
bugfix: when a user coroutine died with an error, its parent coroutine's status was still 'normal'. thanks jinglong for reporting this issue.
|
2012-08-29 12:17:38 -07:00 |
|
agentzh (Yichun Zhang)
|
62d67ad2f2
|
optimize: removed a duplidate ngx_memzero invocation in ngx_http_lua_reset_ctx. thanks jinglong.
|
2012-08-29 11:59:54 -07:00 |
|
agentzh (Yichun Zhang)
|
35c1fcd04e
|
docs: removed the note for the issues regarding "normal" state user coroutines in coroutine.resume() and coroutine.status() because we have already fixed them.
|
2012-08-28 16:23:33 -07:00 |
|
agentzh (Yichun Zhang)
|
50f20351b6
|
merged "master" into "thread".
|
2012-08-28 16:15:42 -07:00 |
|
agentzh (Yichun Zhang)
|
5cb6e9fa52
|
refactor: now every lua coroutine takes its own context state for its nginx-side activities. bugfix: coroutine.status() returned "suspended" for "normal" coroutines. bugfix: coroutine.resume() did not return an error immediately when operating on "normal" coroutines. feature: added new dtrace static probes http-lua-user-coroutine-yield and http-lua-entry-coroutine-yield.
|
2012-08-28 16:04:08 -07:00 |
|
agentzh (Yichun Zhang)
|
a819a192e7
|
docs: documented the new shdict:flush_expired() method. bumped the version to 0.6.3. also mentioned lua-resty-dns in the SEE ALSO section.
|
2012-08-25 23:55:22 -07:00 |
|
agentzh (Yichun Zhang)
|
f09867d915
|
feature: implemented shdict:flush_expired(max_count) for flushing out expired items up to max_count (or unlimited when max_count == 0). thanks Brian Akins for the patch.
|
2012-08-24 21:00:19 -07:00 |
|
agentzh (Yichun Zhang)
|
b771643dba
|
bumped version number to 0.6.2.
|
2012-08-22 15:25:54 -07:00 |