yahns 1.14.1 → 1.18.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +5 -5
- data/.document +2 -0
- data/.gitignore +0 -1
- data/.olddoc.yml +3 -2
- data/Documentation/GNUmakefile +1 -1
- data/Documentation/design_notes.txt +6 -3
- data/Documentation/yahns-rackup.pod +7 -3
- data/Documentation/yahns.pod +1 -1
- data/Documentation/yahns_config.pod +10 -10
- data/GIT-VERSION-FILE +1 -1
- data/GIT-VERSION-GEN +3 -3
- data/HACKING +13 -13
- data/NEWS +982 -829
- data/README +11 -12
- data/Rakefile +121 -5
- data/examples/https_proxy_pass.conf.rb +36 -0
- data/examples/logrotate.conf +1 -1
- data/examples/proxy_pass.ru +11 -0
- data/extras/autoindex.rb +20 -4
- data/extras/exec_cgi.rb +38 -24
- data/extras/proxy_pass.rb +7 -6
- data/extras/try_gzip_static.rb +4 -1
- data/lib/yahns/acceptor.rb +3 -3
- data/lib/yahns/chunk_body.rb +2 -1
- data/lib/yahns/config.rb +10 -5
- data/lib/yahns/daemon.rb +0 -1
- data/lib/yahns/http_client.rb +28 -18
- data/lib/yahns/http_response.rb +3 -4
- data/lib/yahns/openssl_client.rb +33 -11
- data/lib/yahns/proxy_http_response.rb +3 -1
- data/lib/yahns/proxy_pass.rb +68 -10
- data/lib/yahns/queue_epoll.rb +4 -0
- data/lib/yahns/queue_kqueue.rb +0 -6
- data/lib/yahns/queue_quitter_pipe.rb +4 -1
- data/lib/yahns/rackup_handler.rb +3 -7
- data/lib/yahns/server.rb +47 -27
- data/lib/yahns/server_mp.rb +3 -4
- data/lib/yahns/sigevent_efd.rb +0 -1
- data/lib/yahns/sigevent_pipe.rb +13 -6
- data/lib/yahns/socket_helper.rb +1 -1
- data/lib/yahns/stream_input.rb +3 -2
- data/lib/yahns/tee_input.rb +1 -3
- data/lib/yahns/version.rb +1 -1
- data/lib/yahns/wbuf.rb +10 -3
- data/lib/yahns/worker.rb +8 -0
- data/lib/yahns.rb +12 -7
- data/man/yahns-rackup.1 +17 -17
- data/man/yahns.1 +11 -15
- data/man/yahns_config.5 +31 -31
- data/test/helper.rb +6 -2
- data/test/server_helper.rb +20 -5
- data/test/test_bin.rb +33 -30
- data/test/test_config.rb +2 -2
- data/test/test_extras_exec_cgi.rb +24 -1
- data/test/test_extras_try_gzip_static.rb +1 -1
- data/test/test_mt_accept.rb +0 -2
- data/test/test_proxy_pass.rb +1 -2
- data/test/test_proxy_pass_no_buffering.rb +1 -1
- data/test/test_rack_env.rb +58 -0
- data/test/test_serve_static.rb +0 -1
- data/test/test_server.rb +1 -4
- data/test/test_ssl.rb +2 -0
- data/test/test_unix_socket.rb +1 -3
- data/test/test_wbuf.rb +1 -1
- data/yahns.gemspec +8 -5
- metadata +12 -9
data/NEWS
CHANGED
@@ -1,833 +1,986 @@
|
|
1
|
-
|
2
|
-
|
3
|
-
|
4
|
-
|
5
|
-
|
6
|
-
|
7
|
-
|
8
|
-
|
9
|
-
|
10
|
-
|
11
|
-
|
12
|
-
|
13
|
-
|
14
|
-
|
15
|
-
|
16
|
-
|
17
|
-
|
18
|
-
|
19
|
-
|
20
|
-
|
21
|
-
|
22
|
-
|
23
|
-
|
24
|
-
|
25
|
-
|
26
|
-
|
27
|
-
|
28
|
-
|
29
|
-
|
30
|
-
|
31
|
-
|
32
|
-
|
33
|
-
|
34
|
-
|
35
|
-
|
36
|
-
|
37
|
-
|
38
|
-
|
39
|
-
|
40
|
-
|
41
|
-
|
42
|
-
|
43
|
-
|
44
|
-
|
45
|
-
|
46
|
-
|
47
|
-
|
48
|
-
|
49
|
-
|
50
|
-
|
51
|
-
|
52
|
-
|
53
|
-
|
54
|
-
|
55
|
-
|
56
|
-
|
57
|
-
|
58
|
-
|
59
|
-
|
60
|
-
|
61
|
-
|
62
|
-
|
63
|
-
|
64
|
-
|
65
|
-
|
66
|
-
|
67
|
-
|
68
|
-
|
69
|
-
|
70
|
-
|
71
|
-
|
72
|
-
|
73
|
-
|
74
|
-
|
75
|
-
|
76
|
-
|
77
|
-
|
78
|
-
|
79
|
-
|
80
|
-
|
81
|
-
|
82
|
-
|
83
|
-
|
84
|
-
|
85
|
-
|
86
|
-
|
87
|
-
|
88
|
-
|
89
|
-
|
90
|
-
|
91
|
-
|
92
|
-
|
93
|
-
|
94
|
-
|
95
|
-
|
96
|
-
|
97
|
-
|
98
|
-
|
99
|
-
|
100
|
-
|
101
|
-
|
102
|
-
|
103
|
-
|
104
|
-
|
105
|
-
|
106
|
-
|
107
|
-
|
108
|
-
|
109
|
-
|
110
|
-
|
111
|
-
|
112
|
-
|
113
|
-
|
114
|
-
|
115
|
-
|
116
|
-
|
117
|
-
|
118
|
-
|
119
|
-
|
120
|
-
|
121
|
-
|
122
|
-
|
123
|
-
|
124
|
-
|
125
|
-
|
126
|
-
|
127
|
-
|
128
|
-
|
129
|
-
|
130
|
-
|
131
|
-
|
132
|
-
|
133
|
-
|
134
|
-
|
135
|
-
|
136
|
-
|
137
|
-
|
138
|
-
|
139
|
-
|
140
|
-
|
141
|
-
|
142
|
-
|
143
|
-
|
144
|
-
|
145
|
-
|
146
|
-
|
147
|
-
|
148
|
-
|
149
|
-
|
150
|
-
|
151
|
-
|
152
|
-
|
153
|
-
|
154
|
-
|
155
|
-
|
156
|
-
|
157
|
-
|
158
|
-
|
159
|
-
|
160
|
-
|
161
|
-
|
162
|
-
|
163
|
-
|
164
|
-
|
165
|
-
|
166
|
-
|
167
|
-
|
168
|
-
|
169
|
-
|
170
|
-
|
171
|
-
|
172
|
-
|
173
|
-
|
174
|
-
|
175
|
-
|
176
|
-
|
177
|
-
|
178
|
-
|
179
|
-
|
180
|
-
|
181
|
-
|
182
|
-
|
183
|
-
|
184
|
-
|
185
|
-
|
186
|
-
|
187
|
-
|
188
|
-
|
189
|
-
|
190
|
-
|
191
|
-
|
192
|
-
|
1
|
+
yahns 1.18.0 / 2021-10-09
|
2
|
+
-------------------------
|
3
|
+
|
4
|
+
8 changes since 1.17.0 (2019-04-22):
|
5
|
+
|
6
|
+
worker: workaround old F_SETPIPE_SZ bug
|
7
|
+
doc: favor File.read over IO.read to ease review
|
8
|
+
proxy_pass: document as a public API
|
9
|
+
doc: include Yahns/ directory on website
|
10
|
+
extras/try_gzip_static: set "Vary: Accept-Encoding" on gzip
|
11
|
+
do not sleep if signals are pending
|
12
|
+
server: workaround Linux v5.5..v5.13 epoll bug
|
13
|
+
gemspec: allow unicorn 6.x
|
14
|
+
|
15
|
+
yahns 1.17.0 - Earth Day release / 2019-04-22
|
16
|
+
---------------------------------------------
|
17
|
+
|
18
|
+
This releases includes a few kqueue-related bugfixes from
|
19
|
+
Lin Jen-Shin <godfat@godfat.org>. There's also some
|
20
|
+
shutdown_timeout-related bugfixes and some cleanups to
|
21
|
+
reduce dependencies on kgio.
|
22
|
+
|
23
|
+
extras/exec_cgi and extras/autoindex both got some minor
|
24
|
+
updates; the latter of which defaults to a dark scheme to save
|
25
|
+
power on OLED and CRT displays (it is Earth Day, after all :>)
|
26
|
+
|
27
|
+
Eric Wong (25):
|
28
|
+
doc: fix git URLs in README and HACKING
|
29
|
+
http_client: do not warn on do_pread overreach
|
30
|
+
remove IO#close_on_exec= calls
|
31
|
+
yahns/worker: shrink pipe under Linux
|
32
|
+
test/test_extras_exec_cgi.rb: improve test reliability
|
33
|
+
extras/exec_cgi: remove kgio dependency
|
34
|
+
extras/exec_cgi: update copyright year and use SPDX
|
35
|
+
extras/exec_cgi: @body_tip is always set
|
36
|
+
extras/exec_cgi: support Process.spawn options (e.g. RLIMIT_*)
|
37
|
+
server_mp: favor "Kernel#select" instead of "IO.select"
|
38
|
+
server_mp: remove redundant srand call
|
39
|
+
extras/exec_cgi: remove unecessary .freeze calls
|
40
|
+
server: respect shutdown_timeout
|
41
|
+
server: extra shutdown timeout debugging
|
42
|
+
wbuf: use IO#write directly in Ruby 2.5+ for writev
|
43
|
+
worker: rely on F_SETPIPE_SZ to set mininum value
|
44
|
+
remove calls to kgio_wait_* able
|
45
|
+
do not shutdown until listeners are all stopped
|
46
|
+
README: remove SMTP-based subscription info
|
47
|
+
sigevent_*: fixup non-eventfd path
|
48
|
+
README: note HTTPS support for HTTP/1.1
|
49
|
+
extras/autoindex: use dark colors by default
|
50
|
+
doc: update homepage to point to cgit install
|
51
|
+
doc: update wording to avoid "mailing list"
|
52
|
+
doc: remove inaccurate comment about "GNU-ism"
|
53
|
+
|
54
|
+
Lin Jen-Shin (2):
|
55
|
+
Add QueueQuitter#closed? to queue_quitter_pipe.rb
|
56
|
+
@srv.shutdown could raise Errno::ENOTCONN
|
57
|
+
|
58
|
+
yahns 1.16.0 / 2018-08-06
|
59
|
+
-------------------------
|
60
|
+
|
61
|
+
This release fixes warnings for users combining Rack::Deflater
|
62
|
+
with HTTPS support:
|
63
|
+
https://yhbt.net/yahns-public/20180714005630.11812-1-e@80x24.org/
|
64
|
+
|
65
|
+
yahns-rackup(1) users may not specify "-O listen=inherit"
|
66
|
+
when spawning from systemd. The "listen" directive may
|
67
|
+
be omitted entirely from the yahns-config(5) files.
|
68
|
+
|
69
|
+
A few other cleanups and nothing interesting, otherwise.
|
70
|
+
|
71
|
+
20 yawn-worthy changes since v1.15.0 (2017-03-23):
|
72
|
+
config: more descriptive variable name
|
73
|
+
proxy_pass: comment explaining what rack.hijack calls
|
74
|
+
USR2 upgrades may use Process.spawn for vfork
|
75
|
+
avoid Thread#[] and Thread#[]= across threads
|
76
|
+
gemspec: declare Ruby 2.0+ dependency, here
|
77
|
+
test_bin: SO_KEEPALIVE value only needs to be true
|
78
|
+
rackup_handler: remove unnecessary branch
|
79
|
+
test_bin: use RbConfig.ruby for non-standard names
|
80
|
+
allow omitting specifying socket name when inheriting listen socket
|
81
|
+
server: fix incomplete comment about SSLContext#setup
|
82
|
+
test_rack_env: additional test for Rack environment
|
83
|
+
test: allow setting TAIL env to watch error logs
|
84
|
+
fix some unused variables
|
85
|
+
tests: thread-safety fixes
|
86
|
+
test/server_helper: describe reason for termination
|
87
|
+
test/test_ssl: set SSLContext#security_level=0
|
88
|
+
doc: https:// URLs instead of git://
|
89
|
+
http_client: clear backtrace on "wrong version number" in OpenSSL
|
90
|
+
openssl_client: do not attempt writes after SystemCallError
|
91
|
+
use IO#pread if available in Ruby 2.5
|
92
|
+
|
93
|
+
*ZZZZZZ*
|
94
|
+
|
95
|
+
yahns 1.15.0 / 2017-03-23
|
96
|
+
-------------------------
|
97
|
+
|
98
|
+
Minor changes to reduce allocations and simplify our code
|
99
|
+
and dependencies. Nothing particularly interesting unless
|
100
|
+
you're the type of person who appreciates brake pedals
|
101
|
+
in with holes drilled in them for weight reduction.
|
102
|
+
|
103
|
+
11 changes since 1.14.1 (2016-12-14):
|
104
|
+
stream_input: avoid allocation for common #read case
|
105
|
+
proxy_pass: add a note about the instability of this
|
106
|
+
tee_input: simplify conditional for writing to temporary file
|
107
|
+
proxy_http_response: reduce memory pressure from larger headers
|
108
|
+
http_response: make response headers eligible for GC, sooner
|
109
|
+
update more referenced URLs to be HTTPS
|
110
|
+
chunk_body: nodoc this internal class
|
111
|
+
Revert "use olddoc 1.1.0 for generating NEWS + NEWS.atom.xml"
|
112
|
+
gemspec: stop advertising "private" email address
|
113
|
+
doc: design_notes: we do not use EPOLLEXCLUSIVE
|
114
|
+
README: update with disclaimer about subscription
|
115
|
+
|
116
|
+
Ovt punatrf pbzvat...
|
117
|
+
|
118
|
+
yahns 1.14.1 - bugfixes only / 2016-12-14
|
119
|
+
-----------------------------------------
|
120
|
+
|
121
|
+
This release avoids confusing HTTP/1.1 clients with a
|
122
|
+
"Transfer-Encoding: chunked" header on bodyless responses.
|
123
|
+
The header was causing "curl -T" to wait indefinitely for
|
124
|
+
a response body after the server sent a 204.
|
125
|
+
This regression was introduced with autochunk introduced
|
126
|
+
with yahns 1.13(*).
|
127
|
+
|
128
|
+
rack.hijack was also broken for ancient "HTTP/0.9" requests,
|
129
|
+
and now fixed.
|
130
|
+
|
131
|
+
4 changes since 1.14.0:
|
132
|
+
|
133
|
+
queue_*: fix outdated comments
|
134
|
+
http_response: support rack.hijack on HTTP/0.9 responses
|
135
|
+
response: do not set chunked header on bodyless responses
|
136
|
+
proxy_pass: do not chunk on bodyless upstream responses
|
137
|
+
|
138
|
+
(*) https://yhbt.net/yahns-public/20160805-yahns-1.13.0-released@lucky13/
|
139
|
+
|
140
|
+
yahns 1.14.0 - removing undefined behavior / 2016-11-14
|
141
|
+
-------------------------------------------------------
|
142
|
+
|
143
|
+
There's minor feature removals for undocumented and
|
144
|
+
undefined features and behavior which are unlikely to
|
145
|
+
affect anybody unless they serve HTTPS.
|
146
|
+
|
147
|
+
Our website is now self-hosted with HTTPS support (HTTP remains
|
148
|
+
supported for legacy systems):
|
149
|
+
|
150
|
+
https://yhbt.net/yahns/README
|
151
|
+
|
152
|
+
See git history at git://yhbt.net/yahns.git for full details.
|
153
|
+
|
154
|
+
openssl_client: avoid undefined SSL_write behavior
|
155
|
+
move website to https://yhbt.net/yahns/
|
156
|
+
stream_file: remove #to_io support from responses
|
157
|
+
response: only stream "file" responses on known length
|
158
|
+
response: fixup compile error
|
159
|
+
req_res: do not send 502 on catchall error if response buffered
|
160
|
+
|
161
|
+
yahns 1.13.0 - some user-visible improvements... / 2016-08-05
|
162
|
+
-------------------------------------------------------------
|
163
|
+
|
164
|
+
And probably a billion new regressions!
|
165
|
+
|
166
|
+
yahns now allows users to skip the Rack::Head, Rack::Chunked and
|
167
|
+
Rack::ContentLength middlewares to ease migrating from/to other
|
168
|
+
real-world Rack HTTP servers. Most notably, our chunked
|
169
|
+
encoding implementation is a bit faster than Rack::Chunked by
|
170
|
+
taking advantage of the writev(2) syscall:
|
171
|
+
|
172
|
+
https://yhbt.net/yahns-public/20160803031906.14553-4-e@80x24.org/
|
173
|
+
|
174
|
+
There's also rack 2.x fixes in the test case and extras/ section
|
175
|
+
(these incompatibilities did not affect existing users unless
|
176
|
+
they use the wonky extras/ section).
|
177
|
+
|
178
|
+
There's also some graceful shutdown fixes, the process title is
|
179
|
+
now changed to display the number of live FDs.
|
180
|
+
|
181
|
+
Of course, there's the usual round of documentation improvements
|
182
|
+
which are systemd and OpenSSL setup-related this time around.
|
183
|
+
|
184
|
+
However, the majority of changes (proxy_*, wbuf_lite), affect
|
185
|
+
currently-unadvertised functionality which is subject to removal
|
186
|
+
or incompatible config changes. However, they are used to serve
|
187
|
+
our mailing list archives at:
|
188
|
+
|
189
|
+
https://yhbt.net/yahns-public/
|
190
|
+
|
191
|
+
49 changes since yahns 1.12.5:
|
192
|
+
proxy_pass: simplify writing request bodies upstream
|
193
|
+
proxy_pass: hoist out proxy_res_headers method
|
194
|
+
proxy_pass: simplify proxy_http_response
|
195
|
+
proxy_pass: split out body and trailer reading in response
|
196
|
+
proxy_pass: trim down proxy_response_finish, too
|
197
|
+
proxy_pass: split out req_res into a separate file
|
198
|
+
proxy_pass: fix resumes after complete buffering is unblocked
|
199
|
+
proxy_pass: X-Forwarded-For appends to existing list
|
200
|
+
proxy_pass: pass entire object to proxy_http_response
|
201
|
+
proxy_pass: support "proxy_buffering: false"
|
202
|
+
proxy_pass: remove unnecessary rescue
|
203
|
+
req_res: store proxy_pass object here, instead
|
204
|
+
proxy_pass: redo "proxy_buffering: false"
|
205
|
+
wbuf: remove needless "busy" parameter
|
206
|
+
Merge branch 'maint'
|
207
|
+
extras/try_gzip_static: do not show backtrace on syscall errors
|
208
|
+
wbuf: remove tmpdir parameter
|
209
|
+
wbuf_lite: fix write retries for OpenSSL sockets
|
210
|
+
test_proxy_pass_no_buffering: fix racy test
|
211
|
+
queue_*: check for closed IO objects
|
212
|
+
cleanup graceful shutdown handling
|
213
|
+
proxy_pass: more descriptive error messages
|
214
|
+
proxy_pass: fix HTTP/1.0 backends on EOF w/o buffering
|
215
|
+
wbuf_common: reset offset counter when done
|
216
|
+
extras/try_gzip_static: resolve symlinks
|
217
|
+
test_ssl: remove unnecessary priv_key DH parameter
|
218
|
+
openssl_client: wrap shutdown for graceful termination
|
219
|
+
proxy_pass: keep trailer buffer on blocked client writes
|
220
|
+
proxy_pass: avoid TOCTTOU race when unbuffering, too
|
221
|
+
proxy_pass: avoid accessing logger in env after hijacking
|
222
|
+
proxy_pass: avoid stuck responses in "proxy_buffering: false"
|
223
|
+
extras: include status messages in responses
|
224
|
+
update init and add systemd examples
|
225
|
+
test_proxy_pass_no_buffering: exclude rb/ru files, too
|
226
|
+
wbuf_lite: use StringIO instead of TmpIO
|
227
|
+
wbuf_lite: truncate StringIO when done
|
228
|
+
wbuf_lite: prevent clobbering responses
|
229
|
+
wbuf_lite: unify EOF error handling
|
230
|
+
wbuf_lite: reset sf_offset/sf_count consistently
|
231
|
+
wbuf_lite: clear @busy flag when re-arming
|
232
|
+
http_response: drop bodies for non-compliant responses
|
233
|
+
fix rack 2.x compatibility bugs
|
234
|
+
doc: add session cache usage to OpenSSL example
|
235
|
+
test: skip some buffering tests on non-default values
|
236
|
+
response: drop clients after HTTP responses of unknown length
|
237
|
+
response: reduce stack overhead for parameter passing
|
238
|
+
response: support auto-chunking for HTTP/1.1
|
239
|
+
Revert "document Rack::Chunked/ContentLength semi-requirements"
|
240
|
+
extras/exec_cgi: fix for HTTPoxy vulnerability
|
241
|
+
|
242
|
+
yahns 1.12.5 - proxy_pass + rack.hijack fixes / 2016-06-05
|
243
|
+
----------------------------------------------------------
|
244
|
+
|
245
|
+
Hopefully the last of the 1.12.x series, this release
|
246
|
+
fixes a few minor bugs mainly needed for testing.
|
247
|
+
|
248
|
+
No upgrade should be necessary for non-proxy_pass users.
|
249
|
+
|
250
|
+
4 changes since v1.12.4 from the "maint" branch at
|
251
|
+
git://yhbt.net/yahns.git
|
252
|
+
|
253
|
+
http_client: set state to :ignore before hijack callback
|
254
|
+
test/test_client_expire: fix for high RLIMIT_NOFILE
|
255
|
+
proxy_pass: do not chunk HTTP/1.0 with keep-alive
|
256
|
+
proxy_pass: X-Forwarded-For appends to existing list
|
257
|
+
|
258
|
+
lib/yahns/http_client.rb | 6 +++---
|
259
|
+
lib/yahns/proxy_http_response.rb | 8 ++++++--
|
260
|
+
lib/yahns/proxy_pass.rb | 5 ++++-
|
261
|
+
test/test_client_expire.rb | 13 +++++++++++--
|
262
|
+
test/test_proxy_pass.rb | 10 ++++++++++
|
263
|
+
5 files changed, 34 insertions(+), 8 deletions(-)
|
264
|
+
|
265
|
+
Note: the current "master" branch (at commit 5e211ea003d2)
|
266
|
+
includes refactorings and new features not included in this
|
267
|
+
release.
|
268
|
+
|
269
|
+
yahns 1.12.4 - rack.hijack and proxy_pass bugfixes / 2016-05-02
|
270
|
+
---------------------------------------------------------------
|
271
|
+
|
272
|
+
This release fixes some resource leaks in uncommonly used parts
|
273
|
+
of yahns as well as including some documentation improvements.
|
274
|
+
No need to upgrade unless you rely on rack.hijack for responses
|
275
|
+
or use the (currently-undocumented) proxy_pass module(*).
|
276
|
+
|
277
|
+
9 non-merge changes since 1.12.3:
|
278
|
+
|
279
|
+
proxy_pass: honor wbuf_persist when ending response
|
280
|
+
proxy_http_response: fix non-terminated fast responses, too
|
281
|
+
test_proxy_pass: test for auto chunking on 1.0 backends
|
282
|
+
wbuf: drop persistence if writing to client fails
|
283
|
+
proxy_http_response: cleanup: avoid redundant setting of "alive"
|
284
|
+
proxy_http_response: do not persist upstream on slow clients
|
285
|
+
proxy_pass: drop resources immediately on errors
|
286
|
+
document Rack::Chunked/ContentLength semi-requirements
|
287
|
+
extras/exec_cgi: document cgit example
|
288
|
+
|
289
|
+
Documentation/yahns-rackup.pod | 10 ++++++++++
|
290
|
+
GIT-VERSION-GEN | 2 +-
|
291
|
+
examples/yahns_rack_basic.conf.rb | 6 ++++++
|
292
|
+
extras/exec_cgi.rb | 8 ++++++++
|
293
|
+
lib/yahns/proxy_http_response.rb | 40 ++++++++++++++++++++++-----------------
|
294
|
+
lib/yahns/proxy_pass.rb | 5 +++--
|
295
|
+
lib/yahns/wbuf_common.rb | 1 +
|
296
|
+
test/test_proxy_pass.rb | 15 +++++++++++++++
|
297
|
+
8 files changed, 67 insertions(+), 20 deletions(-)
|
298
|
+
|
299
|
+
(*) 1.13.0 will include refactoring in proxy_pass and possibly
|
300
|
+
documenting it as stable-enough-for-public use:
|
301
|
+
|
193
302
|
https://yhbt.net/yahns-public/20160220081619.GA10850@dcvr.yhbt.net/t/
|
194
303
|
|
195
|
-
|
304
|
+
yahns 1.12.3 - more fixes and doc updates / 2016-04-08
|
305
|
+
------------------------------------------------------
|
306
|
+
|
307
|
+
This release only contains two changes since v1.12.2:
|
308
|
+
|
309
|
+
* proxy_http_response: workaround non-terminated backends
|
310
|
+
|
311
|
+
The first is a fix for the undocumented and unstable "proxy_pass"
|
312
|
+
feature:
|
313
|
+
https://yhbt.net/yahns-public/20160406062556.10988-1-e@80x24.org/t/
|
314
|
+
|
315
|
+
However, I'm considering supporting proxy_pass as a stable API:
|
316
|
+
https://yhbt.net/yahns-public/20160220081619.GA10850@dcvr.yhbt.net/t/
|
317
|
+
|
318
|
+
* doc: recommend "verify_mode: OpenSSL::SSL::VERIFY_NONE"
|
319
|
+
|
320
|
+
Only a documentation change prompted by the discovery that
|
321
|
+
some browsers/platforms will try to prompt users for client
|
322
|
+
certs:
|
323
|
+
https://yhbt.net/yahns-public/20160316003434.GA14791@dcvr.yhbt.net/t/
|
196
324
|
|
197
|
-
|
198
|
-
|
199
|
-
certs:
|
200
|
-
https://yhbt.net/yahns-public/20160316003434.GA14791@dcvr.yhbt.net/t/
|
201
|
-
|
202
|
-
=== yahns 1.12.2 - minor doc and TLS fixes / 2016-03-01 01:55 UTC
|
203
|
-
|
204
|
-
This release ensures OpenSSL::SSL::SSLContext#session_id_context
|
205
|
-
is always set for OpenSSL users. It won't overwrite existing
|
206
|
-
settings, but setting it to a random value is necessary to
|
207
|
-
ensure clients do not get aborted connections when attempting to
|
208
|
-
use a session cache.
|
209
|
-
|
210
|
-
No need to actually upgrade if you're on 1.12.1, you may add the
|
211
|
-
following to your yahns_config(5) file where
|
212
|
-
OpenSSL::SSL::SSLContext is configured:
|
213
|
-
|
214
|
-
# recommended, not required. This sets safer defaults
|
215
|
-
# provided by Ruby on top of what OpenSSL gives:
|
216
|
-
ssl_ctx.set_params
|
217
|
-
|
218
|
-
# required, and done by default in v1.12.2:
|
219
|
-
ssl_ctx.session_id_context ||= OpenSSL::Random.random_bytes(32)
|
220
|
-
|
221
|
-
yahns gives you full control of of how OpenSSL::SSL::SSLContext is
|
222
|
-
configured. To avoid bugs, yahns only ensures
|
223
|
-
OpenSSL::SSL::SSLContext#session_id_context is set (if not previously
|
224
|
-
set by the user) and calls OpenSSL::SSL::SSLContext#setup before
|
225
|
-
spawning threads to avoid race conditions. yahns itself does not and
|
226
|
-
will not enforce any opinion on the compatibility/performance/security
|
227
|
-
trade-offs regarding TLS configuration.
|
228
|
-
|
229
|
-
Note: keep in mind using an SSL session cache may be less useful
|
230
|
-
with yahns because HTTP/1.1 persistent connections may live
|
231
|
-
forever :)
|
232
|
-
|
233
|
-
3 bug/doc fixes on top of v1.12.1:
|
234
|
-
document OpenSSL::SSL::SSLContext#set_params use
|
235
|
-
ssl: ensure is session_id_context is always set
|
236
|
-
test/*: fix mktmpdir usage for 1.9.3
|
237
|
-
|
238
|
-
=== yahns 1.12.1 - more TLS fixes / 2016-02-22 00:39 UTC
|
239
|
-
|
240
|
-
Most notably release fixes TLS output buffering for large
|
241
|
-
responses to slow clients. For Rack HTTPS users,
|
242
|
-
env['SERVER_PORT'] also defaults to 443 properly unless the
|
243
|
-
Host: request header specifies differently.
|
244
|
-
|
245
|
-
Also, the extras/autoindex change is to make our own directory
|
246
|
-
listing look nicer as we use Let's Encrypt and don't want to
|
247
|
-
waste space listing ".well-known/" directory contents on:
|
248
|
-
|
249
|
-
https://yahns.yhbt.net/
|
250
|
-
|
251
|
-
Yes, we really do care how our homepage looks!
|
252
|
-
|
253
|
-
6 changes since v1.12.1:
|
254
|
-
extras/autoindex: support hiding dotfiles
|
255
|
-
fix output buffering with SSL_write
|
256
|
-
https: ensure SERVER_PORT defaults to 443
|
257
|
-
test_ssl: check SERVER_PORT when parsed from Host: header
|
258
|
-
doc: mention kqueue/kevent alongside epoll
|
259
|
-
doc: more minor updates
|
260
|
-
|
261
|
-
=== yahns 1.12.0 - TLS fixes and more! / 2016-02-14 22:30 UTC
|
262
|
-
|
263
|
-
Most notably, serving static files over HTTPS did not work
|
264
|
-
before this release with the "sendfile" gem installed. The
|
265
|
-
yahns_config(5) manpage is also updated with an example for
|
266
|
-
using OpenSSL::SSL::SSLContext objects. Users of
|
267
|
-
Rack::Request#scheme and env['rack.url_scheme'] should see
|
268
|
-
"https" properly set for HTTPS connections.
|
269
|
-
|
270
|
-
There's also a bunch of internal tweaks like taking advantage of
|
271
|
-
the file-level frozen_string_literal: directive in 2.3 and
|
272
|
-
explicitly clearing short-lived string buffers
|
273
|
-
|
274
|
-
TLS support is still in its early stages, but I'm experimenting
|
275
|
-
with Let's Encrypt (via getssl[1]) and hosting https://YHBT.net/
|
276
|
-
on it.
|
277
|
-
|
278
|
-
For now, I suggest using a separate yahns instance (with a
|
279
|
-
different master process) to avoid any potential data leaks
|
280
|
-
between HTTPS and HTTP instances. In the future, it may be
|
281
|
-
possible to isolate HTTPS from HTTP at the worker process level.
|
282
|
-
Supporting GnuTLS (alongside OpenSSL) may be in our future, too.
|
283
|
-
|
284
|
-
To paraphrase the warning in http://www.postfix.org/TLS_README.html
|
285
|
-
(which was written before Heartbleed):
|
286
|
-
|
287
|
-
WARNING
|
288
|
-
|
289
|
-
By turning on TLS support in yahns, you not only get the
|
290
|
-
ability to encrypt traffic and to authenticate remote
|
291
|
-
clients. You also turn on thousands and thousands of
|
292
|
-
lines of OpenSSL library code. Assuming that OpenSSL is
|
293
|
-
written as carefully as Eric's own code, every 1000 lines
|
294
|
-
introduce one additional bug into yahns.
|
295
|
-
|
296
|
-
I'm not nearly as careful with yahns as Wietse is with postfix,
|
297
|
-
either.
|
298
|
-
|
299
|
-
20 changes since v1.11.0:
|
300
|
-
README: updates for kqueue
|
301
|
-
add .gitattributes for Ruby method detection
|
302
|
-
nodoc internals
|
303
|
-
enable frozen_string_literal for Ruby 2.3+
|
304
|
-
copyright updates for 2016
|
305
|
-
extras/exec_cgi: fix frozen string error on slow responses
|
306
|
-
avoid StringIO#binmode for the next few years
|
307
|
-
use String#clear for short-lived buffers we create
|
308
|
-
gemspec: make rack a development dependency
|
309
|
-
build: install-gem forced to "--local" domain
|
310
|
-
acceptor: all subclasses of TCPServer use TCP_INFO
|
311
|
-
properly emulate sendfile for OpenSSL sockets
|
312
|
-
avoid race conditions in OpenSSL::SSL::SSLContext#setup
|
313
|
-
set HTTPS and rack.url_scheme in Rack env as appropriate
|
314
|
-
proxy_pass: pass X-Forwarded-Proto through
|
315
|
-
doc: switch to perlpod (from pandoc-flavored Markdown)
|
316
|
-
doc: trim down documentation slightly
|
317
|
-
doc: document ssl_ctx for "listen" directive
|
318
|
-
doc: various doc and linkification improvements
|
319
|
-
http_context: reduce constant lookup + bytecode
|
320
|
-
|
321
|
-
[1] git clone https://github.com/srvrco/getssl.git
|
322
|
-
|
323
|
-
=== yahns 1.11.0 - more minor updates / 2015-12-13 02:20 UTC
|
324
|
-
|
325
|
-
There's some minor test case fixes and documentation updates.
|
326
|
-
|
327
|
-
For OpenSSL users running the Ruby 2.3.0 preview releases,
|
328
|
-
we now use `exception: false' for for accept_nonblock to reduce
|
329
|
-
exceptions. Non-SSL I/O still uses kgio for now.
|
330
|
-
|
331
|
-
6 changes since 1.10.0:
|
332
|
-
wbuf: fix typo in bug check for sendfile gem
|
333
|
-
test_wbuf: deal with proper zero-copy for Unix sockets
|
334
|
-
README: clarify and update copyright year
|
335
|
-
doc: document and reference sd_listen_fds(3) behavior
|
336
|
-
test_server: fix race condition in hooks test
|
337
|
-
openssl_client: use `exception: false' for accept_nonblock
|
338
|
-
|
339
|
-
=== yahns 1.10.0 - minor updates / 2015-11-01 09:08 UTC
|
340
|
-
|
341
|
-
* test_client_expire: favor Process.spawn over fork+exec
|
342
|
-
* extras/try_gzip_static: require readability on HEAD requests
|
343
|
-
* proxy_pass: quiet down ECONNRESET and EPIPE, too
|
344
|
-
* copyright updates
|
345
|
-
* update various comments
|
346
|
-
* proxy_http_response: use frozen string literal optimization
|
347
|
-
* README: add link to NNTP archive endpoint
|
348
|
-
* use olddoc 1.1.0 for generating NEWS + NEWS.atom.xml
|
349
|
-
* gemspec fixes for various RubyGems versions
|
350
|
-
* bytecode golfing to cut a few bytes of memory
|
351
|
-
* simplify sd_listen_fds emulation
|
352
|
-
* gemspec: allow compatibility with unicorn 5
|
353
|
-
|
354
|
-
=== yahns 1.9.0 - minor updates / 2015-07-21 20:29 UTC
|
355
|
-
|
356
|
-
This release improves socket inheritance support. TCP socket
|
357
|
-
options are now applied to inherited sockets. We also emulate
|
358
|
-
the sd_listen_fds function to allow inheriting sockets from
|
359
|
-
systemd.
|
360
|
-
|
361
|
-
HTTP status strings are now generated dynamically, allowing
|
362
|
-
applications to modify Rack::Utils::HTTP_STATUS_CODES to
|
363
|
-
apply changes in the Rack response. Unfortunately, this leads
|
364
|
-
to minor (likely unnoticeable) performance regressions.
|
365
|
-
|
366
|
-
However, our code is not optimized for Ruby 2.2+, so users on
|
367
|
-
the latest released Ruby will benefit from reduced inline cache
|
368
|
-
and constant lookups as we reduced our constant footprint.
|
369
|
-
Expect further minor performance regressions if you are running
|
370
|
-
Ruby 2.2 and earlier.
|
371
|
-
|
372
|
-
For Ruby 2.2 users, overall performance should be largely
|
373
|
-
unchanged from 1.7.0 to 1.8.0
|
374
|
-
|
375
|
-
shortlog of changes since 1.7.0:
|
376
|
-
|
377
|
-
* use opt_str_freeze for Hash#delete
|
378
|
-
* test/helper: warn atomically
|
379
|
-
* generate response status strings dynamically
|
380
|
-
* reduce constants and optimize for Ruby 2.2+
|
381
|
-
* http_response: reduce bytecode size
|
382
|
-
* apply TCP socket options on inherited sockets
|
383
|
-
* test/test_rack_hijack.rb: try to increase test reliability
|
384
|
-
* emulate sd_listen_fds for systemd support
|
385
|
-
* test/test_rack_hijack: ensure proper ordering of log messages
|
386
|
-
|
387
|
-
=== yahns 1.8.0 - minor updates / 2015-06-11 19:54 UTC
|
388
|
-
|
389
|
-
Most notably, the Rack response body is now closed during rack.hijack.
|
390
|
-
|
391
|
-
Middlewares such as Rack::Lock (used by Rails) break badly unless
|
392
|
-
the response body is closed on hijack, so we will close it to follow
|
393
|
-
the lead of other popular Rack servers.
|
394
|
-
|
395
|
-
While it's unclear if there's anybody using rack.hijack besides
|
396
|
-
yahns/proxy_pass we'll try to emulate the behavior of other servers
|
397
|
-
as much as possible.
|
398
|
-
|
399
|
-
ref: https://github.com/ngauthier/tubesock/issues/10
|
400
|
-
|
401
|
-
We'll also support SIGWINCH if not daemonized
|
402
|
-
|
403
|
-
This has no effect for the (default) single process case with
|
404
|
-
no master/worker relationship as that does not support SIGWINCH.
|
405
|
-
|
406
|
-
Some process managers such as foreman and daemontools rely on
|
407
|
-
yahnsnot daemonizing, but we still want to be able to process
|
408
|
-
SIGWINCH in that case.
|
409
|
-
|
410
|
-
stdout and stderr may be redirected to a pipe (for cronolog or
|
411
|
-
similar process), so those are less likely to be attached to a TTY
|
412
|
-
than stdin. This also allows users to process SIGWINCH when running
|
413
|
-
inside a regular terminal if they redirect stdin to /dev/null.
|
414
|
-
|
415
|
-
This follows unicorn commit a6077391bb62d0b13016084b0eea36b987afe8f0
|
416
|
-
Thanks to Dan Moore for suggesting it on the unicorn list.
|
417
|
-
|
418
|
-
A few more minor changes, more memory reduction changes coming...
|
419
|
-
|
420
|
-
* proxy_pass: no point in closing StringIO
|
421
|
-
* proxy_pass: allow filtering or overriding response headers
|
422
|
-
* support SIGWINCH even if not daemonized
|
423
|
-
* use Unicorn::HttpParser#response_start_sent accessor
|
424
|
-
* reduce inline constant cache overheads
|
425
|
-
* proxy_pass: skip tests if kcar is missing
|
426
|
-
* ensure body is closed during hijack
|
427
|
-
|
428
|
-
=== yahns 1.7.0 - rack.hijack improvements and more / 2015-05-11 01:38 UTC
|
429
|
-
|
430
|
-
Nothing really significant, so there's no need to upgrade if
|
431
|
-
you're not affected by the minor fixes and changes in this
|
432
|
-
release.
|
433
|
-
|
434
|
-
For all users, LoadError and SyntaxError exceptions are now
|
435
|
-
logged and non-fatal within worker threads serving application
|
436
|
-
code. Thanks to Lin Jen-Shin <godfat@godfat.org> for bringing
|
437
|
-
this up on the mailing list.
|
438
|
-
|
439
|
-
Additionally, temporary files buffered to the filesystem will
|
440
|
-
now support the Rack::TempfileReaper middleware in rack 1.6+
|
441
|
-
|
442
|
-
For rack.hijack users, there are some changes and improvements.
|
443
|
-
rack.hijack should return a usable IO-like object for SSL users,
|
444
|
-
now. The rack.input object is no longer closed on hijacking,
|
445
|
-
allowing apps to continue using buffered input after hijacking.
|
446
|
-
There is also a bugfix for the rare apps which hijack requests
|
447
|
-
after emitting 100-continue responses.
|
448
|
-
|
449
|
-
Note: there is also a work-in-progress and under-documented
|
450
|
-
asynchronous Yayns::ProxyPass Rack app which uses rack.hijack
|
451
|
-
internally. This will allow yahns to act as a fully-buffering
|
452
|
-
reverse proxy to upstream servers which cannot handle slow
|
453
|
-
clients. Yahns::ProxyPass NOT production-ready as of this
|
454
|
-
release. The old, synchronous extras/proxy_pass.rb code
|
455
|
-
remains usable.
|
456
|
-
|
457
|
-
There's also the usual round of minor code bloat reduction.
|
458
|
-
|
459
|
-
=== yahns 1.6.0 - reduced allocations and bugfixes / 2015-03-09 09:33 UTC
|
460
|
-
|
461
|
-
This release fixes a bug where previously-configured-but-now-removed
|
462
|
-
listeners were inherited across USR2 upgrades are not shutdown
|
463
|
-
immediately in the child.
|
464
|
-
|
465
|
-
There are also minor reductions in allocations which can save a few
|
466
|
-
hundred bytes statically and also whenever write buffering is necessary
|
467
|
-
for large responses.
|
468
|
-
|
469
|
-
Some minor documentation updates improvements in extras, too.
|
470
|
-
|
471
|
-
shortlog of changes since 1.5.0:
|
472
|
-
README: add link to mailing list archives
|
473
|
-
test_ssl: factor out server SSLContext creation
|
474
|
-
doc: add design_notes document
|
475
|
-
reduce File::Stat object allocations
|
476
|
-
update comments about wbuf_close return values
|
477
|
-
wbuf: lazily (re)create temporary file
|
478
|
-
fix compatibility with unicorn.git
|
479
|
-
skip tests requiring String#b on 1.9.3
|
480
|
-
use the monotonic clock under Ruby 2.1+
|
481
|
-
favor Class.new for method-less classes
|
482
|
-
extras/proxy_pass: save memory in String#split arg
|
483
|
-
extras/proxy_pass: do not name unused variable
|
484
|
-
extras/proxy_pass: log exceptions leading to 502
|
485
|
-
extras/proxy_pass: flesh out upload support + tests
|
486
|
-
acceptor: close inherited-but-unneeded sockets
|
487
|
-
|
488
|
-
See the git repository for more: git clone git://yhbt.net/yahns
|
489
|
-
|
490
|
-
=== yahns 1.5.0 - initial OpenSSL support and bugfixes / 2014-12-21 02:22 UTC
|
491
|
-
|
492
|
-
This release adds basic OpenSSL support for HTTPS connections.
|
493
|
-
|
494
|
-
Users must supply a OpenSSL::SSL::SSLContext object which yahns will use
|
495
|
-
as-is. yahns will only support HTTPS on Ruby 2.1 and later, as we rely
|
496
|
-
on "exception: false" in the read_nonblock and write_nonblock methods in
|
497
|
-
OpenSSL::SSL::SSLSocket.
|
498
|
-
|
499
|
-
See the Ruby standard library documentation on how to configure
|
500
|
-
OpenSSL::SSL::SSLContext objects to pass to the yahns "listen" directive
|
501
|
-
Editing the yahns config file to use OpenSSL goes something like this:
|
502
|
-
|
503
|
-
require 'openssl' # we will not do this for the user, even
|
504
|
-
ctx = OpenSSL::SSL::SSLContext.new
|
505
|
-
# user must configure ctx here...
|
506
|
-
|
507
|
-
listen 443, ssl_ctx: ctx
|
508
|
-
|
509
|
-
Note: yahns developers are not responsible for bugs in OpenSSL itself
|
510
|
-
or misconfigured SSLContext objects created by users. However, our
|
511
|
-
support of OpenSSL sockets is barely-tested and likely buggy, too.
|
512
|
-
|
513
|
-
Furthermore, the "sendfile" (or "kgio-sendfile") gem is no longer a
|
514
|
-
required dependency. It is currently impossible to use zero-copy
|
515
|
-
system calls with TLS sockets.
|
516
|
-
|
517
|
-
There are also minor cleanups and a bugfix to ensure body#close is
|
518
|
-
called for folks using body#to_path where `body' is the Rack
|
519
|
-
response body. This bug affected logging using the 'clogger' gem
|
520
|
-
when serving static files.
|
521
|
-
|
522
|
-
Shortlog of changes since 1.4.0
|
523
|
-
|
524
|
-
save around 1500 bytes of memory on x86-64
|
525
|
-
http_response: remove arg for Array#join
|
526
|
-
remove unused client_max_header_size config
|
527
|
-
config: use literal symbol array for now
|
528
|
-
http_response: reduce constants for 100 responses
|
529
|
-
favor Array#map! for freshly-split arrays
|
530
|
-
sendfile_compat: remove dependency on pread
|
531
|
-
extras/autoindex: simplify checking non-.gz
|
532
|
-
Rakefile: kill more useless gsub use
|
533
|
-
initial cut at OpenSSL support
|
534
|
-
test/test_ssl: skip test if SSL on older Rubies
|
535
|
-
wbuf_common: close body proxies on sendfile abort
|
536
|
-
bump published Ruby version requirement to 2.0
|
537
|
-
make sendfile an optional dependency
|
538
|
-
openssl_client: ignore SSL_accept errors during negotiation
|
539
|
-
|
540
|
-
Disclaimer: the yahns project does not and will never endorse
|
541
|
-
any commercial entities, including certificate authorities.
|
542
|
-
|
543
|
-
Shpx Nhgubevgl.
|
544
|
-
|
545
|
-
=== yahns 1.4.0 / 2014-11-16 11:05 UTC
|
546
|
-
|
547
|
-
Minor internal cleanups and bugfixes.
|
548
|
-
|
549
|
-
* http_client: clear some internal ivars on rack.hijack
|
550
|
-
* README: minor updates
|
551
|
-
* extras/autoindex: do not link beyond top-level
|
552
|
-
* extras/autoindex: fix misnamed variable
|
553
|
-
* extras/autoindex: integrate with TryGzipStatic
|
554
|
-
* queue: remove TODO comment for kqueue
|
555
|
-
* switch to require_relative for internal requires
|
556
|
-
* README: add OpenSSL exception
|
557
|
-
* stream_input: favor str.clear instead of str.replace('')
|
558
|
-
* http_response: skip Date header case-insensitively
|
559
|
-
|
560
|
-
=== yahns 1.3.1 - fix large response corruption / 2014-07-16 20:03 UTC
|
561
|
-
|
562
|
-
This release contains a major bug fix noticeable on FreeBSD VMs, but
|
563
|
-
should affect Linux systems making large responses which require
|
564
|
-
output buffering, too.
|
565
|
-
|
566
|
-
* test_server: avoid multiple workers for dead parent check
|
567
|
-
* wbuf: avoid corrupted large responses with zero-copy sendfile
|
568
|
-
|
569
|
-
=== yahns 1.3.0 - new mailing list + fixups / 2014-05-12 01:44 UTC
|
570
|
-
|
571
|
-
Since RubyForge is going away, this release updates the
|
572
|
-
documentation to point to the new public-inbox and mailing list
|
573
|
-
at at:
|
574
|
-
|
575
|
-
yahns-public@yhbt.net
|
576
|
-
(no subscription required, plain text only)
|
577
|
-
|
578
|
-
ssoma[1] git archives: git://yhbt.net/yahns-public
|
579
|
-
browser-friendly archives: http://yhbt.net/yahns-public/
|
580
|
-
Traditional mailing list subscription is possible by sending
|
581
|
-
a blank email to: yahns-public+subscribe@yhbt.net
|
582
|
-
|
583
|
-
This also reworks master->worker signalling to use a pipe,
|
584
|
-
avoiding problems with libraries which do not respond well
|
585
|
-
to being interrupted by signals. There are also minor
|
586
|
-
bugfixes and cleanups, see "git log -p" for full details.
|
587
|
-
|
588
|
-
[1] http://ssoma.public-inbox.org/
|
589
|
-
|
590
|
-
=== yahns 1.2.0 - preliminary kqueue/FreeBSD support / 2014-03-15 07:43 UTC
|
591
|
-
|
592
|
-
This release now depends on "kgio-sendfile", a (hopefully temporary)
|
593
|
-
fork of the original sendfile gem for mainline ruby trunk
|
594
|
-
compatibility and a (probably correct) FreeBSD-related bugfix.
|
595
|
-
|
596
|
-
kqueue/FreeBSD support is considered highly experimental. Of course;
|
597
|
-
you should never rely on anything in production unless you can get bugs
|
598
|
-
fixed in every part of your stack; even the kernel. yahns (ab)uses
|
599
|
-
kqueue and epoll in uncommon ways, so you may encounter subtle kernel
|
600
|
-
bugs along the way.
|
601
|
-
|
602
|
-
Because yahns has been self-hosting its own website for months without
|
603
|
-
crashes or major problems (BORING! :P), I've decided to start hosting the
|
604
|
-
yahns website <http://yahns.YHBT.net/README> with ruby trunk (currently
|
605
|
-
r45341).
|
606
|
-
|
607
|
-
yahns - dangerous by design (and sleepy!)
|
608
|
-
|
609
|
-
=== yahns 1.1.0 - MOAR SLEEPY / 2014-02-04 22:14 UTC
|
610
|
-
|
611
|
-
On responses with a known, non-zero Content-Length, yahns will now
|
612
|
-
automatically use the MSG_MORE flag when sending HTTP headers. This
|
613
|
-
reduces TCP packet transmits and allows clients to wake up and read
|
614
|
-
one less time.
|
615
|
-
|
616
|
-
This is similar to the TCP_NOPUSH/TCP_CORK functionality of other
|
617
|
-
servers, but requires no additional syscalls. It is only supported
|
618
|
-
on the Linux kernel, however (but yahns is currently epoll-only
|
619
|
-
with kqueue support on the horizon).
|
620
|
-
|
621
|
-
Eric Wong (5):
|
622
|
-
quiet down EHOSTUNREACH errors
|
623
|
-
http_response: use kgio_syssend with MSG_MORE
|
624
|
-
load yahns/version file
|
625
|
-
socket_helper: remove SO_REUSEPORT define for untested arches
|
626
|
-
response: do not use MSG_MORE on empty bodies
|
627
|
-
|
628
|
-
=== yahns 1.0.0 / 2014-01-02 01:37 UTC
|
629
|
-
|
630
|
-
Mostly fixes for the extras/* stuff running on yahns.YHBT.net and
|
631
|
-
documentation/packaging fixes. Anyways, this is hosting the
|
632
|
-
yahns website for a while now (running rack.git) without any
|
633
|
-
major issues.
|
634
|
-
|
635
|
-
Also, the mailing list will need to move/change because Rubyforge is
|
636
|
-
going away. Everything I said for the unicorn ML applies to this list,
|
637
|
-
too: http://mid.gmane.org/20131217015614.GB1125@dcvr.yhbt.net
|
638
|
-
|
639
|
-
=== yahns 0.0.3 - more bugfixes, still working! / 2013-11-10 21:42 UTC
|
640
|
-
|
641
|
-
There are several important bugfixes around graceful shutdown.
|
642
|
-
The shutdown_timeout directive should work more consistently
|
643
|
-
and work even if an app is streaming/trickling slowly to
|
644
|
-
a client.
|
645
|
-
|
646
|
-
This also plugs an odd FD leak in response buffering.
|
647
|
-
|
648
|
-
There's also some internal comments/documentation for folks
|
649
|
-
reviewing the strange internals of yahns.
|
650
|
-
|
651
|
-
Otherwise, it seems to be capable of serving its own website,
|
652
|
-
http://yahns.yhbt.net/README quite well.
|
653
|
-
|
654
|
-
Rack application authors merely need to write code as if they have a
|
655
|
-
gun to their head if they expect to keep code running on yahns.
|
656
|
-
Again, yahns is extremely intolerant of bugs in the applications
|
657
|
-
it hosts. Otherwise I'm comfortable in the stability of yahns
|
658
|
-
itself.
|
659
|
-
|
660
|
-
Eric Wong (17):
|
661
|
-
http_client: do not dump backtrace on ETIMEDOUT
|
662
|
-
ensure we close response body if buffering caught up
|
663
|
-
http_response: reorder wbuf_maybe on successful early flush
|
664
|
-
wbuf: document reasoning for the design of these clases
|
665
|
-
build: improve NEWS generation, add atom feed
|
666
|
-
exec_cgi: release pipe/pid sooner
|
667
|
-
SIGWINCH works after SIGUSR2 upgrades
|
668
|
-
shutdown_timeout works around long-running response/apps
|
669
|
-
doc: recommend worker_processes if the app uses SIGCHLD
|
670
|
-
fdmap: simplify IO expiry interface
|
671
|
-
fdmap: document + fix for level-trigger
|
672
|
-
queue_epoll: document epoll concurrency caveats
|
673
|
-
doc: caution users against disabling buffering
|
674
|
-
queue: eliminate :wait_readwrite
|
675
|
-
test: exec_cgi test uses worker_processes
|
676
|
-
test_server: remove unneccessary IO#wait call
|
677
|
-
test: workaround sockets not being binary on rbx
|
678
|
-
|
679
|
-
=== yahns 0.0.2 - barely working! / 2013-11-06 03:24 UTC
|
680
|
-
|
681
|
-
Lots of bugfixes and tweaks, but everything appears to mostly work
|
682
|
-
for Rack and HTTP. We are self-hosting our site:
|
683
|
-
|
684
|
-
http://yahns.yhbt.net/README
|
685
|
-
(And if you can't access it, I screwed something up!)
|
686
|
-
|
687
|
-
There is a yahns-rackup(1) wrapper for quick-starting Rack
|
688
|
-
applications without having to read any documentation,
|
689
|
-
as it works exactly like rackup(1) (just leave "-s/--server"
|
690
|
-
out)
|
691
|
-
|
692
|
-
New manpages: yahns(1), yahns_config(5), yahns-rackup(1)
|
693
|
-
|
694
|
-
Eric Wong (110):
|
695
|
-
test/helper: correctly handle exit code in test runner
|
696
|
-
log: workaround atomicity issues for stdio vs non-stdio descriptors
|
697
|
-
tests: improve output to show RUBY_DESCRIPTION and full path
|
698
|
-
tests: support disabling parallelization env
|
699
|
-
test_output_buffering: workaround a rbx bug
|
700
|
-
socket_helper: do not log sizes on rbx for now
|
701
|
-
config: do not set cloexec on stdout/stderr
|
702
|
-
test_wbuf: avoid floating point arg to read_nonblock
|
703
|
-
rework shutdown for systems w/o rb_thread_fd_close
|
704
|
-
SIGUSR2 uses fork + exec again instead of Process.spawn
|
705
|
-
wip for EADDRINUSE failure
|
706
|
-
test_bin: shutdown socket when not inheriting
|
707
|
-
README: update with support status for MRI/RBX
|
708
|
-
server: skip killing acceptors on rbx for now
|
709
|
-
server (cleanup): avoid interning word for log message
|
710
|
-
server: abort loudly if we have old threads running
|
711
|
-
reinstate retry delay for binding new listeners
|
712
|
-
test/helper: Dir.mktmpdir works without blocks
|
713
|
-
rework acceptor thread shutdown (again)
|
714
|
-
server (minor cleanup): use Symbol#to_proc
|
715
|
-
test_server: remove skipped multi-process balance test
|
716
|
-
tests: do not lose exceptions on quit timeouts
|
717
|
-
tests: enforce close_on_exec on all client sockets
|
718
|
-
server: switch abort to raise on BUG
|
719
|
-
config: implement atfork handlers
|
720
|
-
test/server_helper: fix undefined variable
|
721
|
-
favor client timeout if lower than desperate timeout
|
722
|
-
tests: enable checks for desperate client expiry
|
723
|
-
implement + test Expect: 100-continue handling
|
724
|
-
quitter: save one syscall and implement for non-eventfd
|
725
|
-
enforce FD_CLOEXEC on all pipes, including tests
|
726
|
-
test_expect_100: cleanup unused var
|
727
|
-
http_client: reduce the size of the yahns_step method
|
728
|
-
output_buffering handles odd EAGAIN cases
|
729
|
-
fix output buffer corner-case for static files
|
730
|
-
tests: increase mkserver use to reduce LoC
|
731
|
-
fix SIGCHLD w/o workers + PID file renames
|
732
|
-
test_client_expire: disable output buffering in test
|
733
|
-
StreamFile and TmpIO attempt expiry on EMFILE/ENFILE
|
734
|
-
server: avoid metaclass for acceptors
|
735
|
-
support SO_REUSEPORT on new listeners (:reuseport)
|
736
|
-
doc: ignore RDoc for all internal classes
|
737
|
-
rack: leave RACK_ENV unset by us
|
738
|
-
associate private/anonymous queues correctly
|
739
|
-
queue_epoll: remove check for closed descriptor
|
740
|
-
config: disallow defining new, named queues inside app
|
741
|
-
doc: preliminary manpages for yahns(1), yahns_config(5)
|
742
|
-
config: comment for atfork_* hook definitions
|
743
|
-
implement user switching
|
744
|
-
configurator: validate :reuseport for boolean-ess
|
745
|
-
config: working_directory is only called at top-level
|
746
|
-
server: fix out-of-date comment regarding bind/pid order
|
747
|
-
config: reject negative float for client_expire_ratio
|
748
|
-
lower client_body_buffer_size to 8K (from 114K)
|
749
|
-
implement client_timeout for streaming inputs
|
750
|
-
gem: install manpages in the RubyGems package
|
751
|
-
stream_input: use thread-local rbuf to avoid some garbage
|
752
|
-
test output_buffering with hijacked responses
|
753
|
-
test_input: close client when we're done with it
|
754
|
-
allow multiple blocking threads per listen socket
|
755
|
-
acceptor: account for inheriting dead descriptors
|
756
|
-
server: fix incorrect receiver of method call
|
757
|
-
socket_helper: account for undefined options
|
758
|
-
test for binding Unix stream sockets
|
759
|
-
Rack hijack issues EPOLL_CTL_DEL
|
760
|
-
config: raise ArgumentError for consistency
|
761
|
-
tests for SIGTTIN and SIGTTOU
|
762
|
-
use Hash#each instead of Hash#each_pair
|
763
|
-
add test for working_directory config parameter
|
764
|
-
test_unix_socket: remember to close IO when done
|
765
|
-
test for overriding rack.errors destination
|
766
|
-
test_server: improve working_directory test robustness
|
767
|
-
fdmap: prevent possible/theoretical race
|
768
|
-
implement shutdown_timeout and expiry on graceful shutdown
|
769
|
-
doc: fix client_timeout documentation in yahns_config
|
770
|
-
implement before_exec hook
|
771
|
-
comment to explain YAHNS_FD env
|
772
|
-
remove "worker_" prefix from atfork hooks
|
773
|
-
remove arity enforcement for atfork_* hooks
|
774
|
-
allow atfork_* hooks inside app blocks for ease-of-management
|
775
|
-
doc: disambiguate threads: option for listen directive
|
776
|
-
yahns-rackup launcher
|
777
|
-
enable client expiry for non-TCP sockets
|
778
|
-
config: allow Float for timeouts
|
779
|
-
tests: more intelligent waiting for output buffering
|
780
|
-
doc: add HACKING/INSTALL docs + minitest dependency
|
781
|
-
wbuf: bypass buffering if buffers are caught up
|
782
|
-
wbuf: reset FS (sendfile) buffer if caught up
|
783
|
-
wbuf: only enable bypass if we've truncated
|
784
|
-
input and output buffers support tmpdir: arguments
|
785
|
-
tests: further speed up output buffering test
|
786
|
-
test trysendfile usage with UNIX sockets
|
787
|
-
client_expire_generic: drop kgio_write wrapper
|
788
|
-
remove unnecessary map(&:to_s) before Array#join
|
789
|
-
account for truncated/expanded static files
|
790
|
-
http: do not drop Content-Range from response headers
|
791
|
-
worker-less server should not waitpid indiscriminately
|
792
|
-
stream_file: only close FDs we opened ourselves
|
793
|
-
tests: smaller buffer for big header test
|
794
|
-
add extras for exec_cgi and try_gzip_static
|
795
|
-
http_response: fix app-controlled close + test
|
796
|
-
examples: flesh out the example configs a bit
|
797
|
-
README: add disclaimer
|
798
|
-
server: improve shutdown messages
|
799
|
-
extras: add README
|
800
|
-
extras: add autoindex module
|
801
|
-
extras: add proxy_pass Rack app
|
802
|
-
rackup_handler: fix ordering of working_directory
|
803
|
-
worker: avoid double SIGQUIT on unexpected parent death
|
804
|
-
server: log error instead of raising for leftover socks
|
805
|
-
|
806
|
-
=== yahns 0.0.1 - many small fixes and test coverage / 2013-10-20 01:00 UTC
|
807
|
-
|
808
|
-
Just a bunch of improvements found while running tests.
|
809
|
-
It's still incomplete and missing a bunch of features from
|
810
|
-
existing servers, but maybe it works...
|
811
|
-
|
812
|
-
Clone git://yhbt.net/yahns.git for full details
|
813
|
-
|
814
|
-
Eric Wong (18):
|
815
|
-
test/helper: fix __covmerge race for forked processes
|
816
|
-
wire up client_max_body_size limits
|
817
|
-
fix and test Rack hijack support
|
818
|
-
SIGUSR2 handling uses Process.spawn + tests
|
819
|
-
fix USR1 log reopening when using worker_processes
|
820
|
-
test_bin: add additional tests for SIGHUP
|
821
|
-
test_server: skip test_mp_balance for now
|
822
|
-
test/server_helper: dump entire log on errors
|
823
|
-
ensure we stop all threads at exit
|
824
|
-
GNUmakefile: avoid calling exit in test-mt
|
825
|
-
test_bin: set close-on-exec for Ruby 1.9.3 compatibility
|
826
|
-
test_server: bigger delays for graceful shutdown test
|
827
|
-
tests: disable $-w on 1.9.3 to quiet down warnings
|
828
|
-
test/helper: prevent minitest at_exit from running in children
|
829
|
-
recheck IO#closed? on thread pools after a short delay
|
830
|
-
test_config: isolate directories with logs
|
831
|
-
test_reopen_logs: workaround timing problem with worker_processes
|
832
|
-
set close-on-exec on all long-lived descriptors
|
325
|
+
yahns 1.12.2 - minor doc and TLS fixes / 2016-03-01
|
326
|
+
---------------------------------------------------
|
833
327
|
|
328
|
+
This release ensures OpenSSL::SSL::SSLContext#session_id_context
|
329
|
+
is always set for OpenSSL users. It won't overwrite existing
|
330
|
+
settings, but setting it to a random value is necessary to
|
331
|
+
ensure clients do not get aborted connections when attempting to
|
332
|
+
use a session cache.
|
333
|
+
|
334
|
+
No need to actually upgrade if you're on 1.12.1, you may add the
|
335
|
+
following to your yahns_config(5) file where
|
336
|
+
OpenSSL::SSL::SSLContext is configured:
|
337
|
+
|
338
|
+
# recommended, not required. This sets safer defaults
|
339
|
+
# provided by Ruby on top of what OpenSSL gives:
|
340
|
+
ssl_ctx.set_params
|
341
|
+
|
342
|
+
# required, and done by default in v1.12.2:
|
343
|
+
ssl_ctx.session_id_context ||= OpenSSL::Random.random_bytes(32)
|
344
|
+
|
345
|
+
yahns gives you full control of of how OpenSSL::SSL::SSLContext is
|
346
|
+
configured. To avoid bugs, yahns only ensures
|
347
|
+
OpenSSL::SSL::SSLContext#session_id_context is set (if not previously
|
348
|
+
set by the user) and calls OpenSSL::SSL::SSLContext#setup before
|
349
|
+
spawning threads to avoid race conditions. yahns itself does not and
|
350
|
+
will not enforce any opinion on the compatibility/performance/security
|
351
|
+
trade-offs regarding TLS configuration.
|
352
|
+
|
353
|
+
Note: keep in mind using an SSL session cache may be less useful
|
354
|
+
with yahns because HTTP/1.1 persistent connections may live
|
355
|
+
forever :)
|
356
|
+
|
357
|
+
3 bug/doc fixes on top of v1.12.1:
|
358
|
+
document OpenSSL::SSL::SSLContext#set_params use
|
359
|
+
ssl: ensure is session_id_context is always set
|
360
|
+
test/*: fix mktmpdir usage for 1.9.3
|
361
|
+
|
362
|
+
yahns 1.12.1 - more TLS fixes / 2016-02-22
|
363
|
+
------------------------------------------
|
364
|
+
|
365
|
+
Most notably release fixes TLS output buffering for large
|
366
|
+
responses to slow clients. For Rack HTTPS users,
|
367
|
+
env['SERVER_PORT'] also defaults to 443 properly unless the
|
368
|
+
Host: request header specifies differently.
|
369
|
+
|
370
|
+
Also, the extras/autoindex change is to make our own directory
|
371
|
+
listing look nicer as we use Let's Encrypt and don't want to
|
372
|
+
waste space listing ".well-known/" directory contents on:
|
373
|
+
|
374
|
+
https://yahns.yhbt.net/
|
375
|
+
|
376
|
+
Yes, we really do care how our homepage looks!
|
377
|
+
|
378
|
+
6 changes since v1.12.1:
|
379
|
+
extras/autoindex: support hiding dotfiles
|
380
|
+
fix output buffering with SSL_write
|
381
|
+
https: ensure SERVER_PORT defaults to 443
|
382
|
+
test_ssl: check SERVER_PORT when parsed from Host: header
|
383
|
+
doc: mention kqueue/kevent alongside epoll
|
384
|
+
doc: more minor updates
|
385
|
+
|
386
|
+
yahns 1.12.0 - TLS fixes and more! / 2016-02-14
|
387
|
+
-----------------------------------------------
|
388
|
+
|
389
|
+
Most notably, serving static files over HTTPS did not work
|
390
|
+
before this release with the "sendfile" gem installed. The
|
391
|
+
yahns_config(5) manpage is also updated with an example for
|
392
|
+
using OpenSSL::SSL::SSLContext objects. Users of
|
393
|
+
Rack::Request#scheme and env['rack.url_scheme'] should see
|
394
|
+
"https" properly set for HTTPS connections.
|
395
|
+
|
396
|
+
There's also a bunch of internal tweaks like taking advantage of
|
397
|
+
the file-level frozen_string_literal: directive in 2.3 and
|
398
|
+
explicitly clearing short-lived string buffers
|
399
|
+
|
400
|
+
TLS support is still in its early stages, but I'm experimenting
|
401
|
+
with Let's Encrypt (via getssl[1]) and hosting https://YHBT.net/
|
402
|
+
on it.
|
403
|
+
|
404
|
+
For now, I suggest using a separate yahns instance (with a
|
405
|
+
different master process) to avoid any potential data leaks
|
406
|
+
between HTTPS and HTTP instances. In the future, it may be
|
407
|
+
possible to isolate HTTPS from HTTP at the worker process level.
|
408
|
+
Supporting GnuTLS (alongside OpenSSL) may be in our future, too.
|
409
|
+
|
410
|
+
To paraphrase the warning in http://www.postfix.org/TLS_README.html
|
411
|
+
(which was written before Heartbleed):
|
412
|
+
|
413
|
+
WARNING
|
414
|
+
|
415
|
+
By turning on TLS support in yahns, you not only get the
|
416
|
+
ability to encrypt traffic and to authenticate remote
|
417
|
+
clients. You also turn on thousands and thousands of
|
418
|
+
lines of OpenSSL library code. Assuming that OpenSSL is
|
419
|
+
written as carefully as Eric's own code, every 1000 lines
|
420
|
+
introduce one additional bug into yahns.
|
421
|
+
|
422
|
+
I'm not nearly as careful with yahns as Wietse is with postfix,
|
423
|
+
either.
|
424
|
+
|
425
|
+
20 changes since v1.11.0:
|
426
|
+
README: updates for kqueue
|
427
|
+
add .gitattributes for Ruby method detection
|
428
|
+
nodoc internals
|
429
|
+
enable frozen_string_literal for Ruby 2.3+
|
430
|
+
copyright updates for 2016
|
431
|
+
extras/exec_cgi: fix frozen string error on slow responses
|
432
|
+
avoid StringIO#binmode for the next few years
|
433
|
+
use String#clear for short-lived buffers we create
|
434
|
+
gemspec: make rack a development dependency
|
435
|
+
build: install-gem forced to "--local" domain
|
436
|
+
acceptor: all subclasses of TCPServer use TCP_INFO
|
437
|
+
properly emulate sendfile for OpenSSL sockets
|
438
|
+
avoid race conditions in OpenSSL::SSL::SSLContext#setup
|
439
|
+
set HTTPS and rack.url_scheme in Rack env as appropriate
|
440
|
+
proxy_pass: pass X-Forwarded-Proto through
|
441
|
+
doc: switch to perlpod (from pandoc-flavored Markdown)
|
442
|
+
doc: trim down documentation slightly
|
443
|
+
doc: document ssl_ctx for "listen" directive
|
444
|
+
doc: various doc and linkification improvements
|
445
|
+
http_context: reduce constant lookup + bytecode
|
446
|
+
|
447
|
+
[1] git clone https://github.com/srvrco/getssl.git
|
448
|
+
|
449
|
+
yahns 1.11.0 - more minor updates / 2015-12-13
|
450
|
+
----------------------------------------------
|
451
|
+
|
452
|
+
There's some minor test case fixes and documentation updates.
|
453
|
+
|
454
|
+
For OpenSSL users running the Ruby 2.3.0 preview releases,
|
455
|
+
we now use `exception: false' for for accept_nonblock to reduce
|
456
|
+
exceptions. Non-SSL I/O still uses kgio for now.
|
457
|
+
|
458
|
+
6 changes since 1.10.0:
|
459
|
+
wbuf: fix typo in bug check for sendfile gem
|
460
|
+
test_wbuf: deal with proper zero-copy for Unix sockets
|
461
|
+
README: clarify and update copyright year
|
462
|
+
doc: document and reference sd_listen_fds(3) behavior
|
463
|
+
test_server: fix race condition in hooks test
|
464
|
+
openssl_client: use `exception: false' for accept_nonblock
|
465
|
+
|
466
|
+
yahns 1.10.0 - minor updates / 2015-11-01
|
467
|
+
-----------------------------------------
|
468
|
+
|
469
|
+
* test_client_expire: favor Process.spawn over fork+exec
|
470
|
+
* extras/try_gzip_static: require readability on HEAD requests
|
471
|
+
* proxy_pass: quiet down ECONNRESET and EPIPE, too
|
472
|
+
* copyright updates
|
473
|
+
* update various comments
|
474
|
+
* proxy_http_response: use frozen string literal optimization
|
475
|
+
* README: add link to NNTP archive endpoint
|
476
|
+
* use olddoc 1.1.0 for generating NEWS + NEWS.atom.xml
|
477
|
+
* gemspec fixes for various RubyGems versions
|
478
|
+
* bytecode golfing to cut a few bytes of memory
|
479
|
+
* simplify sd_listen_fds emulation
|
480
|
+
* gemspec: allow compatibility with unicorn 5
|
481
|
+
|
482
|
+
yahns 1.9.0 - minor updates / 2015-07-21
|
483
|
+
----------------------------------------
|
484
|
+
|
485
|
+
This release improves socket inheritance support. TCP socket
|
486
|
+
options are now applied to inherited sockets. We also emulate
|
487
|
+
the sd_listen_fds function to allow inheriting sockets from
|
488
|
+
systemd.
|
489
|
+
|
490
|
+
HTTP status strings are now generated dynamically, allowing
|
491
|
+
applications to modify Rack::Utils::HTTP_STATUS_CODES to
|
492
|
+
apply changes in the Rack response. Unfortunately, this leads
|
493
|
+
to minor (likely unnoticeable) performance regressions.
|
494
|
+
|
495
|
+
However, our code is not optimized for Ruby 2.2+, so users on
|
496
|
+
the latest released Ruby will benefit from reduced inline cache
|
497
|
+
and constant lookups as we reduced our constant footprint.
|
498
|
+
Expect further minor performance regressions if you are running
|
499
|
+
Ruby 2.2 and earlier.
|
500
|
+
|
501
|
+
For Ruby 2.2 users, overall performance should be largely
|
502
|
+
unchanged from 1.7.0 to 1.8.0
|
503
|
+
|
504
|
+
shortlog of changes since 1.7.0:
|
505
|
+
|
506
|
+
* use opt_str_freeze for Hash#delete
|
507
|
+
* test/helper: warn atomically
|
508
|
+
* generate response status strings dynamically
|
509
|
+
* reduce constants and optimize for Ruby 2.2+
|
510
|
+
* http_response: reduce bytecode size
|
511
|
+
* apply TCP socket options on inherited sockets
|
512
|
+
* test/test_rack_hijack.rb: try to increase test reliability
|
513
|
+
* emulate sd_listen_fds for systemd support
|
514
|
+
* test/test_rack_hijack: ensure proper ordering of log messages
|
515
|
+
|
516
|
+
yahns 1.8.0 - minor updates / 2015-06-11
|
517
|
+
----------------------------------------
|
518
|
+
|
519
|
+
Most notably, the Rack response body is now closed during rack.hijack.
|
520
|
+
|
521
|
+
Middlewares such as Rack::Lock (used by Rails) break badly unless
|
522
|
+
the response body is closed on hijack, so we will close it to follow
|
523
|
+
the lead of other popular Rack servers.
|
524
|
+
|
525
|
+
While it's unclear if there's anybody using rack.hijack besides
|
526
|
+
yahns/proxy_pass we'll try to emulate the behavior of other servers
|
527
|
+
as much as possible.
|
528
|
+
|
529
|
+
ref: https://github.com/ngauthier/tubesock/issues/10
|
530
|
+
|
531
|
+
We'll also support SIGWINCH if not daemonized
|
532
|
+
|
533
|
+
This has no effect for the (default) single process case with
|
534
|
+
no master/worker relationship as that does not support SIGWINCH.
|
535
|
+
|
536
|
+
Some process managers such as foreman and daemontools rely on
|
537
|
+
yahnsnot daemonizing, but we still want to be able to process
|
538
|
+
SIGWINCH in that case.
|
539
|
+
|
540
|
+
stdout and stderr may be redirected to a pipe (for cronolog or
|
541
|
+
similar process), so those are less likely to be attached to a TTY
|
542
|
+
than stdin. This also allows users to process SIGWINCH when running
|
543
|
+
inside a regular terminal if they redirect stdin to /dev/null.
|
544
|
+
|
545
|
+
This follows unicorn commit a6077391bb62d0b13016084b0eea36b987afe8f0
|
546
|
+
Thanks to Dan Moore for suggesting it on the unicorn list.
|
547
|
+
|
548
|
+
A few more minor changes, more memory reduction changes coming...
|
549
|
+
|
550
|
+
* proxy_pass: no point in closing StringIO
|
551
|
+
* proxy_pass: allow filtering or overriding response headers
|
552
|
+
* support SIGWINCH even if not daemonized
|
553
|
+
* use Unicorn::HttpParser#response_start_sent accessor
|
554
|
+
* reduce inline constant cache overheads
|
555
|
+
* proxy_pass: skip tests if kcar is missing
|
556
|
+
* ensure body is closed during hijack
|
557
|
+
|
558
|
+
yahns 1.7.0 - rack.hijack improvements and more / 2015-05-11
|
559
|
+
------------------------------------------------------------
|
560
|
+
|
561
|
+
Nothing really significant, so there's no need to upgrade if
|
562
|
+
you're not affected by the minor fixes and changes in this
|
563
|
+
release.
|
564
|
+
|
565
|
+
For all users, LoadError and SyntaxError exceptions are now
|
566
|
+
logged and non-fatal within worker threads serving application
|
567
|
+
code. Thanks to Lin Jen-Shin <godfat@godfat.org> for bringing
|
568
|
+
this up on the mailing list.
|
569
|
+
|
570
|
+
Additionally, temporary files buffered to the filesystem will
|
571
|
+
now support the Rack::TempfileReaper middleware in rack 1.6+
|
572
|
+
|
573
|
+
For rack.hijack users, there are some changes and improvements.
|
574
|
+
rack.hijack should return a usable IO-like object for SSL users,
|
575
|
+
now. The rack.input object is no longer closed on hijacking,
|
576
|
+
allowing apps to continue using buffered input after hijacking.
|
577
|
+
There is also a bugfix for the rare apps which hijack requests
|
578
|
+
after emitting 100-continue responses.
|
579
|
+
|
580
|
+
Note: there is also a work-in-progress and under-documented
|
581
|
+
asynchronous Yayns::ProxyPass Rack app which uses rack.hijack
|
582
|
+
internally. This will allow yahns to act as a fully-buffering
|
583
|
+
reverse proxy to upstream servers which cannot handle slow
|
584
|
+
clients. Yahns::ProxyPass NOT production-ready as of this
|
585
|
+
release. The old, synchronous extras/proxy_pass.rb code
|
586
|
+
remains usable.
|
587
|
+
|
588
|
+
There's also the usual round of minor code bloat reduction.
|
589
|
+
|
590
|
+
yahns 1.6.0 - reduced allocations and bugfixes / 2015-03-09
|
591
|
+
-----------------------------------------------------------
|
592
|
+
|
593
|
+
This release fixes a bug where previously-configured-but-now-removed
|
594
|
+
listeners were inherited across USR2 upgrades are not shutdown
|
595
|
+
immediately in the child.
|
596
|
+
|
597
|
+
There are also minor reductions in allocations which can save a few
|
598
|
+
hundred bytes statically and also whenever write buffering is necessary
|
599
|
+
for large responses.
|
600
|
+
|
601
|
+
Some minor documentation updates improvements in extras, too.
|
602
|
+
|
603
|
+
shortlog of changes since 1.5.0:
|
604
|
+
README: add link to mailing list archives
|
605
|
+
test_ssl: factor out server SSLContext creation
|
606
|
+
doc: add design_notes document
|
607
|
+
reduce File::Stat object allocations
|
608
|
+
update comments about wbuf_close return values
|
609
|
+
wbuf: lazily (re)create temporary file
|
610
|
+
fix compatibility with unicorn.git
|
611
|
+
skip tests requiring String#b on 1.9.3
|
612
|
+
use the monotonic clock under Ruby 2.1+
|
613
|
+
favor Class.new for method-less classes
|
614
|
+
extras/proxy_pass: save memory in String#split arg
|
615
|
+
extras/proxy_pass: do not name unused variable
|
616
|
+
extras/proxy_pass: log exceptions leading to 502
|
617
|
+
extras/proxy_pass: flesh out upload support + tests
|
618
|
+
acceptor: close inherited-but-unneeded sockets
|
619
|
+
|
620
|
+
See the git repository for more: git clone git://yhbt.net/yahns
|
621
|
+
|
622
|
+
yahns 1.5.0 - initial OpenSSL support and bugfixes / 2014-12-21
|
623
|
+
---------------------------------------------------------------
|
624
|
+
|
625
|
+
This release adds basic OpenSSL support for HTTPS connections.
|
626
|
+
|
627
|
+
Users must supply a OpenSSL::SSL::SSLContext object which yahns will use
|
628
|
+
as-is. yahns will only support HTTPS on Ruby 2.1 and later, as we rely
|
629
|
+
on "exception: false" in the read_nonblock and write_nonblock methods in
|
630
|
+
OpenSSL::SSL::SSLSocket.
|
631
|
+
|
632
|
+
See the Ruby standard library documentation on how to configure
|
633
|
+
OpenSSL::SSL::SSLContext objects to pass to the yahns "listen" directive
|
634
|
+
Editing the yahns config file to use OpenSSL goes something like this:
|
635
|
+
|
636
|
+
require 'openssl' # we will not do this for the user, even
|
637
|
+
ctx = OpenSSL::SSL::SSLContext.new
|
638
|
+
# user must configure ctx here...
|
639
|
+
|
640
|
+
listen 443, ssl_ctx: ctx
|
641
|
+
|
642
|
+
Note: yahns developers are not responsible for bugs in OpenSSL itself
|
643
|
+
or misconfigured SSLContext objects created by users. However, our
|
644
|
+
support of OpenSSL sockets is barely-tested and likely buggy, too.
|
645
|
+
|
646
|
+
Furthermore, the "sendfile" (or "kgio-sendfile") gem is no longer a
|
647
|
+
required dependency. It is currently impossible to use zero-copy
|
648
|
+
system calls with TLS sockets.
|
649
|
+
|
650
|
+
There are also minor cleanups and a bugfix to ensure body#close is
|
651
|
+
called for folks using body#to_path where `body' is the Rack
|
652
|
+
response body. This bug affected logging using the 'clogger' gem
|
653
|
+
when serving static files.
|
654
|
+
|
655
|
+
Shortlog of changes since 1.4.0
|
656
|
+
|
657
|
+
save around 1500 bytes of memory on x86-64
|
658
|
+
http_response: remove arg for Array#join
|
659
|
+
remove unused client_max_header_size config
|
660
|
+
config: use literal symbol array for now
|
661
|
+
http_response: reduce constants for 100 responses
|
662
|
+
favor Array#map! for freshly-split arrays
|
663
|
+
sendfile_compat: remove dependency on pread
|
664
|
+
extras/autoindex: simplify checking non-.gz
|
665
|
+
Rakefile: kill more useless gsub use
|
666
|
+
initial cut at OpenSSL support
|
667
|
+
test/test_ssl: skip test if SSL on older Rubies
|
668
|
+
wbuf_common: close body proxies on sendfile abort
|
669
|
+
bump published Ruby version requirement to 2.0
|
670
|
+
make sendfile an optional dependency
|
671
|
+
openssl_client: ignore SSL_accept errors during negotiation
|
672
|
+
|
673
|
+
Disclaimer: the yahns project does not and will never endorse
|
674
|
+
any commercial entities, including certificate authorities.
|
675
|
+
|
676
|
+
Shpx Nhgubevgl.
|
677
|
+
|
678
|
+
yahns 1.4.0 / 2014-11-16
|
679
|
+
------------------------
|
680
|
+
|
681
|
+
Minor internal cleanups and bugfixes.
|
682
|
+
|
683
|
+
* http_client: clear some internal ivars on rack.hijack
|
684
|
+
* README: minor updates
|
685
|
+
* extras/autoindex: do not link beyond top-level
|
686
|
+
* extras/autoindex: fix misnamed variable
|
687
|
+
* extras/autoindex: integrate with TryGzipStatic
|
688
|
+
* queue: remove TODO comment for kqueue
|
689
|
+
* switch to require_relative for internal requires
|
690
|
+
* README: add OpenSSL exception
|
691
|
+
* stream_input: favor str.clear instead of str.replace('')
|
692
|
+
* http_response: skip Date header case-insensitively
|
693
|
+
|
694
|
+
yahns 1.3.1 - fix large response corruption / 2014-07-16
|
695
|
+
--------------------------------------------------------
|
696
|
+
|
697
|
+
This release contains a major bug fix noticeable on FreeBSD VMs, but
|
698
|
+
should affect Linux systems making large responses which require
|
699
|
+
output buffering, too.
|
700
|
+
|
701
|
+
* test_server: avoid multiple workers for dead parent check
|
702
|
+
* wbuf: avoid corrupted large responses with zero-copy sendfile
|
703
|
+
|
704
|
+
yahns 1.3.0 - new mailing list + fixups / 2014-05-12
|
705
|
+
----------------------------------------------------
|
706
|
+
|
707
|
+
Since RubyForge is going away, this release updates the
|
708
|
+
documentation to point to the new public-inbox and mailing list
|
709
|
+
at at:
|
710
|
+
|
711
|
+
yahns-public@yhbt.net
|
712
|
+
(no subscription required, plain text only)
|
713
|
+
|
714
|
+
ssoma[1] git archives: git://yhbt.net/yahns-public
|
715
|
+
browser-friendly archives: http://yhbt.net/yahns-public/
|
716
|
+
Traditional mailing list subscription is possible by sending
|
717
|
+
a blank email to: yahns-public+subscribe@yhbt.net
|
718
|
+
|
719
|
+
This also reworks master->worker signalling to use a pipe,
|
720
|
+
avoiding problems with libraries which do not respond well
|
721
|
+
to being interrupted by signals. There are also minor
|
722
|
+
bugfixes and cleanups, see "git log -p" for full details.
|
723
|
+
|
724
|
+
[1] http://ssoma.public-inbox.org/
|
725
|
+
|
726
|
+
yahns 1.2.0 - preliminary kqueue/FreeBSD support / 2014-03-15
|
727
|
+
-------------------------------------------------------------
|
728
|
+
|
729
|
+
This release now depends on "kgio-sendfile", a (hopefully temporary)
|
730
|
+
fork of the original sendfile gem for mainline ruby trunk
|
731
|
+
compatibility and a (probably correct) FreeBSD-related bugfix.
|
732
|
+
|
733
|
+
kqueue/FreeBSD support is considered highly experimental. Of course;
|
734
|
+
you should never rely on anything in production unless you can get bugs
|
735
|
+
fixed in every part of your stack; even the kernel. yahns (ab)uses
|
736
|
+
kqueue and epoll in uncommon ways, so you may encounter subtle kernel
|
737
|
+
bugs along the way.
|
738
|
+
|
739
|
+
Because yahns has been self-hosting its own website for months without
|
740
|
+
crashes or major problems (BORING! :P), I've decided to start hosting the
|
741
|
+
yahns website <http://yahns.YHBT.net/README> with ruby trunk (currently
|
742
|
+
r45341).
|
743
|
+
|
744
|
+
yahns - dangerous by design (and sleepy!)
|
745
|
+
|
746
|
+
yahns 1.1.0 - MOAR SLEEPY / 2014-02-04
|
747
|
+
--------------------------------------
|
748
|
+
|
749
|
+
On responses with a known, non-zero Content-Length, yahns will now
|
750
|
+
automatically use the MSG_MORE flag when sending HTTP headers. This
|
751
|
+
reduces TCP packet transmits and allows clients to wake up and read
|
752
|
+
one less time.
|
753
|
+
|
754
|
+
This is similar to the TCP_NOPUSH/TCP_CORK functionality of other
|
755
|
+
servers, but requires no additional syscalls. It is only supported
|
756
|
+
on the Linux kernel, however (but yahns is currently epoll-only
|
757
|
+
with kqueue support on the horizon).
|
758
|
+
|
759
|
+
Eric Wong (5):
|
760
|
+
quiet down EHOSTUNREACH errors
|
761
|
+
http_response: use kgio_syssend with MSG_MORE
|
762
|
+
load yahns/version file
|
763
|
+
socket_helper: remove SO_REUSEPORT define for untested arches
|
764
|
+
response: do not use MSG_MORE on empty bodies
|
765
|
+
|
766
|
+
yahns 1.0.0 / 2014-01-02
|
767
|
+
------------------------
|
768
|
+
|
769
|
+
Mostly fixes for the extras/* stuff running on yahns.YHBT.net and
|
770
|
+
documentation/packaging fixes. Anyways, this is hosting the
|
771
|
+
yahns website for a while now (running rack.git) without any
|
772
|
+
major issues.
|
773
|
+
|
774
|
+
Also, the mailing list will need to move/change because Rubyforge is
|
775
|
+
going away. Everything I said for the unicorn ML applies to this list,
|
776
|
+
too: http://mid.gmane.org/20131217015614.GB1125@dcvr.yhbt.net
|
777
|
+
|
778
|
+
yahns 0.0.3 - more bugfixes, still working! / 2013-11-10
|
779
|
+
--------------------------------------------------------
|
780
|
+
|
781
|
+
There are several important bugfixes around graceful shutdown.
|
782
|
+
The shutdown_timeout directive should work more consistently
|
783
|
+
and work even if an app is streaming/trickling slowly to
|
784
|
+
a client.
|
785
|
+
|
786
|
+
This also plugs an odd FD leak in response buffering.
|
787
|
+
|
788
|
+
There's also some internal comments/documentation for folks
|
789
|
+
reviewing the strange internals of yahns.
|
790
|
+
|
791
|
+
Otherwise, it seems to be capable of serving its own website,
|
792
|
+
http://yahns.yhbt.net/README quite well.
|
793
|
+
|
794
|
+
Rack application authors merely need to write code as if they have a
|
795
|
+
gun to their head if they expect to keep code running on yahns.
|
796
|
+
Again, yahns is extremely intolerant of bugs in the applications
|
797
|
+
it hosts. Otherwise I'm comfortable in the stability of yahns
|
798
|
+
itself.
|
799
|
+
|
800
|
+
Eric Wong (17):
|
801
|
+
http_client: do not dump backtrace on ETIMEDOUT
|
802
|
+
ensure we close response body if buffering caught up
|
803
|
+
http_response: reorder wbuf_maybe on successful early flush
|
804
|
+
wbuf: document reasoning for the design of these clases
|
805
|
+
build: improve NEWS generation, add atom feed
|
806
|
+
exec_cgi: release pipe/pid sooner
|
807
|
+
SIGWINCH works after SIGUSR2 upgrades
|
808
|
+
shutdown_timeout works around long-running response/apps
|
809
|
+
doc: recommend worker_processes if the app uses SIGCHLD
|
810
|
+
fdmap: simplify IO expiry interface
|
811
|
+
fdmap: document + fix for level-trigger
|
812
|
+
queue_epoll: document epoll concurrency caveats
|
813
|
+
doc: caution users against disabling buffering
|
814
|
+
queue: eliminate :wait_readwrite
|
815
|
+
test: exec_cgi test uses worker_processes
|
816
|
+
test_server: remove unneccessary IO#wait call
|
817
|
+
test: workaround sockets not being binary on rbx
|
818
|
+
|
819
|
+
yahns 0.0.2 - barely working! / 2013-11-06
|
820
|
+
------------------------------------------
|
821
|
+
|
822
|
+
Lots of bugfixes and tweaks, but everything appears to mostly work
|
823
|
+
for Rack and HTTP. We are self-hosting our site:
|
824
|
+
|
825
|
+
http://yahns.yhbt.net/README
|
826
|
+
(And if you can't access it, I screwed something up!)
|
827
|
+
|
828
|
+
There is a yahns-rackup(1) wrapper for quick-starting Rack
|
829
|
+
applications without having to read any documentation,
|
830
|
+
as it works exactly like rackup(1) (just leave "-s/--server"
|
831
|
+
out)
|
832
|
+
|
833
|
+
New manpages: yahns(1), yahns_config(5), yahns-rackup(1)
|
834
|
+
|
835
|
+
Eric Wong (110):
|
836
|
+
test/helper: correctly handle exit code in test runner
|
837
|
+
log: workaround atomicity issues for stdio vs non-stdio descriptors
|
838
|
+
tests: improve output to show RUBY_DESCRIPTION and full path
|
839
|
+
tests: support disabling parallelization env
|
840
|
+
test_output_buffering: workaround a rbx bug
|
841
|
+
socket_helper: do not log sizes on rbx for now
|
842
|
+
config: do not set cloexec on stdout/stderr
|
843
|
+
test_wbuf: avoid floating point arg to read_nonblock
|
844
|
+
rework shutdown for systems w/o rb_thread_fd_close
|
845
|
+
SIGUSR2 uses fork + exec again instead of Process.spawn
|
846
|
+
wip for EADDRINUSE failure
|
847
|
+
test_bin: shutdown socket when not inheriting
|
848
|
+
README: update with support status for MRI/RBX
|
849
|
+
server: skip killing acceptors on rbx for now
|
850
|
+
server (cleanup): avoid interning word for log message
|
851
|
+
server: abort loudly if we have old threads running
|
852
|
+
reinstate retry delay for binding new listeners
|
853
|
+
test/helper: Dir.mktmpdir works without blocks
|
854
|
+
rework acceptor thread shutdown (again)
|
855
|
+
server (minor cleanup): use Symbol#to_proc
|
856
|
+
test_server: remove skipped multi-process balance test
|
857
|
+
tests: do not lose exceptions on quit timeouts
|
858
|
+
tests: enforce close_on_exec on all client sockets
|
859
|
+
server: switch abort to raise on BUG
|
860
|
+
config: implement atfork handlers
|
861
|
+
test/server_helper: fix undefined variable
|
862
|
+
favor client timeout if lower than desperate timeout
|
863
|
+
tests: enable checks for desperate client expiry
|
864
|
+
implement + test Expect: 100-continue handling
|
865
|
+
quitter: save one syscall and implement for non-eventfd
|
866
|
+
enforce FD_CLOEXEC on all pipes, including tests
|
867
|
+
test_expect_100: cleanup unused var
|
868
|
+
http_client: reduce the size of the yahns_step method
|
869
|
+
output_buffering handles odd EAGAIN cases
|
870
|
+
fix output buffer corner-case for static files
|
871
|
+
tests: increase mkserver use to reduce LoC
|
872
|
+
fix SIGCHLD w/o workers + PID file renames
|
873
|
+
test_client_expire: disable output buffering in test
|
874
|
+
StreamFile and TmpIO attempt expiry on EMFILE/ENFILE
|
875
|
+
server: avoid metaclass for acceptors
|
876
|
+
support SO_REUSEPORT on new listeners (:reuseport)
|
877
|
+
doc: ignore RDoc for all internal classes
|
878
|
+
rack: leave RACK_ENV unset by us
|
879
|
+
associate private/anonymous queues correctly
|
880
|
+
queue_epoll: remove check for closed descriptor
|
881
|
+
config: disallow defining new, named queues inside app
|
882
|
+
doc: preliminary manpages for yahns(1), yahns_config(5)
|
883
|
+
config: comment for atfork_* hook definitions
|
884
|
+
implement user switching
|
885
|
+
configurator: validate :reuseport for boolean-ess
|
886
|
+
config: working_directory is only called at top-level
|
887
|
+
server: fix out-of-date comment regarding bind/pid order
|
888
|
+
config: reject negative float for client_expire_ratio
|
889
|
+
lower client_body_buffer_size to 8K (from 114K)
|
890
|
+
implement client_timeout for streaming inputs
|
891
|
+
gem: install manpages in the RubyGems package
|
892
|
+
stream_input: use thread-local rbuf to avoid some garbage
|
893
|
+
test output_buffering with hijacked responses
|
894
|
+
test_input: close client when we're done with it
|
895
|
+
allow multiple blocking threads per listen socket
|
896
|
+
acceptor: account for inheriting dead descriptors
|
897
|
+
server: fix incorrect receiver of method call
|
898
|
+
socket_helper: account for undefined options
|
899
|
+
test for binding Unix stream sockets
|
900
|
+
Rack hijack issues EPOLL_CTL_DEL
|
901
|
+
config: raise ArgumentError for consistency
|
902
|
+
tests for SIGTTIN and SIGTTOU
|
903
|
+
use Hash#each instead of Hash#each_pair
|
904
|
+
add test for working_directory config parameter
|
905
|
+
test_unix_socket: remember to close IO when done
|
906
|
+
test for overriding rack.errors destination
|
907
|
+
test_server: improve working_directory test robustness
|
908
|
+
fdmap: prevent possible/theoretical race
|
909
|
+
implement shutdown_timeout and expiry on graceful shutdown
|
910
|
+
doc: fix client_timeout documentation in yahns_config
|
911
|
+
implement before_exec hook
|
912
|
+
comment to explain YAHNS_FD env
|
913
|
+
remove "worker_" prefix from atfork hooks
|
914
|
+
remove arity enforcement for atfork_* hooks
|
915
|
+
allow atfork_* hooks inside app blocks for ease-of-management
|
916
|
+
doc: disambiguate threads: option for listen directive
|
917
|
+
yahns-rackup launcher
|
918
|
+
enable client expiry for non-TCP sockets
|
919
|
+
config: allow Float for timeouts
|
920
|
+
tests: more intelligent waiting for output buffering
|
921
|
+
doc: add HACKING/INSTALL docs + minitest dependency
|
922
|
+
wbuf: bypass buffering if buffers are caught up
|
923
|
+
wbuf: reset FS (sendfile) buffer if caught up
|
924
|
+
wbuf: only enable bypass if we've truncated
|
925
|
+
input and output buffers support tmpdir: arguments
|
926
|
+
tests: further speed up output buffering test
|
927
|
+
test trysendfile usage with UNIX sockets
|
928
|
+
client_expire_generic: drop kgio_write wrapper
|
929
|
+
remove unnecessary map(&:to_s) before Array#join
|
930
|
+
account for truncated/expanded static files
|
931
|
+
http: do not drop Content-Range from response headers
|
932
|
+
worker-less server should not waitpid indiscriminately
|
933
|
+
stream_file: only close FDs we opened ourselves
|
934
|
+
tests: smaller buffer for big header test
|
935
|
+
add extras for exec_cgi and try_gzip_static
|
936
|
+
http_response: fix app-controlled close + test
|
937
|
+
examples: flesh out the example configs a bit
|
938
|
+
README: add disclaimer
|
939
|
+
server: improve shutdown messages
|
940
|
+
extras: add README
|
941
|
+
extras: add autoindex module
|
942
|
+
extras: add proxy_pass Rack app
|
943
|
+
rackup_handler: fix ordering of working_directory
|
944
|
+
worker: avoid double SIGQUIT on unexpected parent death
|
945
|
+
server: log error instead of raising for leftover socks
|
946
|
+
|
947
|
+
yahns 0.0.1 - many small fixes and test coverage / 2013-10-20
|
948
|
+
-------------------------------------------------------------
|
949
|
+
|
950
|
+
Just a bunch of improvements found while running tests.
|
951
|
+
It's still incomplete and missing a bunch of features from
|
952
|
+
existing servers, but maybe it works...
|
953
|
+
|
954
|
+
Clone git://yhbt.net/yahns.git for full details
|
955
|
+
|
956
|
+
Eric Wong (18):
|
957
|
+
test/helper: fix __covmerge race for forked processes
|
958
|
+
wire up client_max_body_size limits
|
959
|
+
fix and test Rack hijack support
|
960
|
+
SIGUSR2 handling uses Process.spawn + tests
|
961
|
+
fix USR1 log reopening when using worker_processes
|
962
|
+
test_bin: add additional tests for SIGHUP
|
963
|
+
test_server: skip test_mp_balance for now
|
964
|
+
test/server_helper: dump entire log on errors
|
965
|
+
ensure we stop all threads at exit
|
966
|
+
GNUmakefile: avoid calling exit in test-mt
|
967
|
+
test_bin: set close-on-exec for Ruby 1.9.3 compatibility
|
968
|
+
test_server: bigger delays for graceful shutdown test
|
969
|
+
tests: disable $-w on 1.9.3 to quiet down warnings
|
970
|
+
test/helper: prevent minitest at_exit from running in children
|
971
|
+
recheck IO#closed? on thread pools after a short delay
|
972
|
+
test_config: isolate directories with logs
|
973
|
+
test_reopen_logs: workaround timing problem with worker_processes
|
974
|
+
set close-on-exec on all long-lived descriptors
|
975
|
+
|
976
|
+
yahns 0.0.0 - hopefully fix installation problems / 2013-10-18
|
977
|
+
--------------------------------------------------------------
|
978
|
+
|
979
|
+
Installing prrerelease versions are strange on RubyGems.
|
980
|
+
Anyways, new project, no risk of breakage, so maybe this
|
981
|
+
works better...
|
982
|
+
|
983
|
+
COPYRIGHT
|
984
|
+
---------
|
985
|
+
Copyright (C) 2013-2017 all contributors <yahns-public@yhbt.net>
|
986
|
+
License: GPL-3.0+ <https://www.gnu.org/licenses/gpl-3.0.txt>
|