ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libev/Changes
(Generate patch)

Comparing libev/Changes (file contents):
Revision 1.99 by root, Wed Jan 7 20:45:52 2009 UTC vs.
Revision 1.366 by root, Sun Mar 8 15:55:43 2020 UTC

1Revision history for libev, a high-performance and full-featured event loop. 1Revision history for libev, a high-performance and full-featured event loop.
2
3TODO: for next ABI/API change, consider moving EV__IOFDSSET into io->fd instead and provide a getter.
4TODO: document EV_TSTAMP_T
5
64.32 (EV only)
7 - the 4.31 timerfd code wrongly changed the priority of the signal
8 fd watcher, which is usually harmless unless signal fds are
9 also used (found via cpan tester service).
10 - the documentation wrongly claimed that user may modify fd and events
11 members in io watchers when the watcher was stopped
12 (found by b_jonas).
13 - new ev_io_modify mutator which changes only the events member,
14 which can be faster. also added ev::io::set (int events) method
15 to ev++.h.
16 - officially allow a zero events mask for io watchers. this should
17 work with older libev versions as well but was not officially
18 allowed before.
19 - do not wake up every minute when timerfd is used to detect timejumps.
20 - do not wake up every minute when periodics are disabled and we have
21 a monotonic clock.
22 - support a lot more "uncommon" compile time configurations,
23 such as ev_embed enabled but ev_timer disabled.
24 - use a start/stop wrapper class to reduce code duplication in
25 ev++.h and make it needlessly more c++-y.
26 - the linux aio backend is no longer compiled in by default.
27 - update to libecb version 0x00010008.
28
294.31 Fri Dec 20 21:58:29 CET 2019
30 - handle backends with minimum wait time a bit better by not
31 waiting in the presence of already-expired timers
32 (behaviour reported by Felipe Gasper).
33 - new feature: use timerfd to detect timejumps quickly,
34 can be disabled with the new EVFLAG_NOTIMERFD loop flag.
35 - document EV_USE_SIGNALFD feature macro.
36
374.30 (EV only)
38 - change non-autoconf test for __kernel_rwf_t by testing
39 LINUX_VERSION_CODE, the most direct test I could find.
40 - fix a bug in the io_uring backend that polled the wrong
41 backend fd, causing it to not work in many cases.
42
434.29 (EV only)
44 - add io uring autoconf and non-autoconf detection.
45 - disable io_uring when some header files are too old.
46
474.28 (EV only)
48 - linuxaio backend resulted in random memory corruption
49 when loop is forked.
50 - linuxaio backend might have tried to cancel an iocb
51 multiple times (was unable to trigger this).
52 - linuxaio backend now employs a generation counter to
53 avoid handling spurious events from cancelled requests.
54 - io_cancel can return EINTR, deal with it. also, assume
55 io_submit also returns EINTR.
56 - fix some other minor bugs in linuxaio backend.
57 - ev_tstamp type can now be overriden by defining EV_TSTAMP_T.
58 - cleanup: replace expect_true/false and noinline by their
59 libecb counterparts.
60 - move syscall infrastructure from ev_linuxaio.c to ev.c.
61 - prepare io_uring integration.
62 - tweak ev_floor.
63 - epoll, poll, win32 Sleep and other places that use millisecond
64 reslution now all try to round up times.
65 - solaris port backend didn't compile.
66 - abstract time constants into their macros, for more flexibility.
67
684.27 Thu Jun 27 22:43:44 CEST 2019
69 - linux aio backend almost completely rewritten to work around its
70 limitations.
71 - linux aio backend now requires linux 4.19+.
72 - epoll backend now mandatory for linux aio backend.
73 - fail assertions more aggressively on invalid fd's detected
74 in the event loop, do not just silently fd_kill in case of
75 user error.
76 - ev_io_start/ev_io_stop now verify the watcher fd using
77 a syscall when EV_VERIFY is 2 or higher.
78
794.26 (EV only)
80 - update to libecb 0x00010006.
81 - new experimental linux aio backend (linux 4.18+).
82 - removed redundant 0-ptr check in ev_once.
83 - updated/extended ev_set_allocator documentation.
84 - replaced EMPTY2 macro by array_needsize_noinit.
85 - minor code cleanups.
86 - epoll backend now uses epoll_create1 also after fork.
87
884.25 Fri Dec 21 07:49:20 CET 2018
89 - INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT
90 (EV_THROW still provided) and now uses noexcept on C++11 or newer.
91 - move the darwin select workaround higher in ev.c, as newer versions of
92 darwin managed to break their broken select even more.
93 - ANDROID => __ANDROID__ (reported by enh@google.com).
94 - disable epoll_create1 on android because it has broken header files
95 and google is unwilling to fix them (reported by enh@google.com).
96 - avoid a minor compilation warning on win32.
97 - c++: remove deprecated dynamic throw() specifications.
98 - c++: improve the (unsupported) bad_loop exception class.
99 - backport perl ev_periodic example to C, untested.
100 - update libecb, biggets change is to include a memory fence
101 in ECB_MEMORY_FENCE_RELEASE on x86/amd64.
102 - minor autoconf/automake modernisation.
103
1044.24 Wed Dec 28 05:19:55 CET 2016
105 - bump version to 4.24, as the release tarball inexplicably
106 didn't have the right version in ev.h, even though the cvs-tagged
107 version did have the right one (reported by Ales Teska).
108
1094.23 Wed Nov 16 18:23:41 CET 2016
110 - move some declarations at the beginning to help certain retarded
111 microsoft compilers, even though their documentation claims
112 otherwise (reported by Ruslan Osmanov).
113
1144.22 Sun Dec 20 22:11:50 CET 2015
115 - when epoll detects unremovable fds in the fd set, rebuild
116 only the epoll descriptor, not the signal pipe, to avoid
117 SIGPIPE in ev_async_send. This doesn't solve it on fork,
118 so document what needs to be done in ev_loop_fork
119 (analyzed by Benjamin Mahler).
120 - remove superfluous sys/timeb.h include on win32
121 (analyzed by Jason Madden).
122 - updated libecb.
123
1244.20 Sat Jun 20 13:01:43 CEST 2015
125 - prefer noexcept over throw () with C++ 11.
126 - update ecb.h due to incompatibilities with c11.
127 - fix a potential aliasing issue when reading and writing
128 watcher callbacks.
129
1304.19 Thu Sep 25 08:18:25 CEST 2014
131 - ev.h wasn't valid C++ anymore, which tripped compilers other than
132 clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
133 C++ doesn't support typedefs for function pointers fully, so the affected
134 declarations have to spell out the types each time.
135 - when not using autoconf, tighten the check for clock_gettime and related
136 functionality.
137
1384.18 Fri Sep 5 17:55:26 CEST 2014
139 - events on files were not always generated properly with the
140 epoll backend (testcase by Assaf Inbal).
141 - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
142 - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
143 - use a reasonable fallback for EV_NSIG instead of erroring out
144 when we can't detect the signal set size.
145 - in the absence of autoconf, do not use the clock syscall
146 on glibc >= 2.17 (avoids the syscall AND -lrt on systems
147 doing clock_gettime in userspace).
148 - ensure extern "C" function pointers are used for externally-visible
149 loop callbacks (not watcher callbacks yet).
150 - (ecb) work around memory barriers and volatile apparently both being
151 broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
152
1534.15 Fri Mar 1 12:04:50 CET 2013
154 - destroying a non-default loop would stop the global waitpid
155 watcher (Denis Bilenko).
156 - queueing pending watchers of higher priority from a watcher now invokes
157 them in a timely fashion (reported by Denis Bilenko).
158 - add throw() to all libev functions that cannot throw exceptions, for
159 further code size decrease when compiling for C++.
160 - add throw () to callbacks that must not throw exceptions (allocator,
161 syserr, loop acquire/release, periodic reschedule cbs).
162 - fix event_base_loop return code, add event_get_callback, event_base_new,
163 event_base_get_method calls to improve libevent 1.x emulation and add
164 some libevent 2.x functionality (based on a patch by Jeff Davey).
165 - add more memory fences to fix a bug reported by Jeff Davey. Better
166 be overfenced than underprotected.
167 - ev_run now returns a boolean status (true meaning watchers are
168 still active).
169 - ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with
170 libev's EV_ERROR (reported by 191919).
171 - (ecb) add memory fence support for xlC (Darin McBride).
172 - (ecb) add memory fence support for gcc-mips (Anton Kirilov).
173 - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber).
174 - work around some kernels losing file descriptors by leaking
175 the kqueue descriptor in the child.
176 - work around linux inotify not reporting IN_ATTRIB changes for directories
177 in many cases.
178 - include sys/syscall.h instead of plain syscall.h.
179 - check for io watcher loops in ev_verify, check for the most
180 common reported usage bug in ev_io_start.
181 - choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
182 - always use WSASend/WSARecv directly on windows, hoping that this
183 works in all cases (unlike read/write/send/recv...).
184 - try to detect signals around a fork faster (test program by
185 Denis Bilenko).
186 - work around recent glibc versions that leak memory in realloc.
187 - rename ev::embed::set to ev::embed::set_embed to avoid clashing
188 the watcher base set (loop) method.
189 - rewrite the async/signal pipe logic to always keep a valid fd, which
190 simplifies (and hopefully correctifies :) the race checking
191 on fork, at the cost of one extra fd.
192 - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
193 inotify-supporting filesystems.
194 - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
195 versions ignore it before
196 (https://bugzilla.redhat.com/show_bug.cgi?id=908096).
197 - add some untested android support.
198 - enum expressions must be of type int (reported by Juan Pablo L).
199
2004.11 Sat Feb 4 19:52:39 CET 2012
201 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
202 was documented already, but not implemented in the repeating case.
203 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
204 - fix a race where the workaround against the epoll fork bugs
205 caused signals to not be handled anymore.
206 - correct backend_fudge for most backends, and implement a windows
207 specific workaround to avoid looping because we call both
208 select and Sleep, both with different time resolutions.
209 - document range and guarantees of ev_sleep.
210 - document reasonable ranges for periodics interval and offset.
211 - rename backend_fudge to backend_mintime to avoid future confusion :)
212 - change the default periodic reschedule function to hopefully be more
213 exact and correct even in corner cases or in the far future.
214 - do not rely on -lm anymore: use it when available but use our
215 own floor () if it is missing. This should make it easier to embed,
216 as no external libraries are required.
217 - strategically import macros from libecb and mark rarely-used functions
218 as cache-cold (saving almost 2k code size on typical amd64 setups).
219 - add Symbols.ev and Symbols.event files, that were missing.
220 - fix backend_mintime value for epoll (was 1/1024, is 1/1000 now).
221 - fix #3 "be smart about timeouts" to not "deadlock" when
222 timeout == now, also improve the section overall.
223 - avoid "AVOIDING FINISHING BEFORE RETURNING" idiom.
224 - support new EV_API_STATIC mode to make all libev symbols
225 static.
226 - supply default CFLAGS of -g -O3 with gcc when original CFLAGS
227 were empty.
228
2294.04 Wed Feb 16 09:01:51 CET 2011
230 - fix two problems in the native win32 backend, where reuse of fd's
231 with different underlying handles caused handles not to be removed
232 or added to the select set (analyzed and tested by Bert Belder).
233 - do no rely on ceil() in ev_e?poll.c.
234 - backport libev to HP-UX versions before 11 v3.
235 - configure did not detect nanosleep and clock_gettime properly when
236 they are available in the libc (as opposed to -lrt).
237
2384.03 Tue Jan 11 14:37:25 CET 2011
239 - officially support polling files with all backends.
240 - support files, /dev/zero etc. the same way as select in the epoll
241 backend, by generating events on our own.
242 - ports backend: work around solaris bug 6874410 and many related ones
243 (EINTR, maybe more), with no performance loss (note that the solaris
244 bug report is actually wrong, reality is far more bizarre and broken
245 than that).
246 - define EV_READ/EV_WRITE as macros in event.h, as some programs use
247 #ifdef to test for them.
248 - new (experimental) function: ev_feed_signal.
249 - new (to become default) EVFLAG_NOSIGMASK flag.
250 - new EVBACKEND_MASK symbol.
251 - updated COMMON IDIOMS SECTION.
252
2534.01 Fri Nov 5 21:51:29 CET 2010
254 - automake fucked it up, apparently, --add-missing -f is not quite enough
255 to make it update its files, so 4.00 didn't install ev++.h and
256 event.h on make install. grrr.
257 - ev_loop(count|depth) didn't return anything (Robin Haberkorn).
258 - change EV_UNDEF to 0xffffffff to silence some overzealous compilers.
259 - use "(libev) " prefix for all libev error messages now.
260
2614.00 Mon Oct 25 12:32:12 CEST 2010
262 - "PORTING FROM LIBEV 3.X TO 4.X" (in ev.pod) is recommended reading.
263 - ev_embed_stop did not correctly stop the watcher (very good
264 testcase by Vladimir Timofeev).
265 - ev_run will now always update the current loop time - it erroneously
266 didn't when idle watchers were active, causing timers not to fire.
267 - fix a bug where a timeout of zero caused the timer not to fire
268 in the libevent emulation (testcase by Péter Szabó).
269 - applied win32 fixes by Michael Lenaghan (also James Mansion).
270 - replace EV_MINIMAL by EV_FEATURES.
271 - prefer EPOLL_CTL_ADD over EPOLL_CTL_MOD in some more cases, as it
272 seems the former is *much* faster than the latter.
273 - linux kernel version detection (for inotify bug workarounds)
274 did not work properly.
275 - reduce the number of spurious wake-ups with the ports backend.
276 - remove dependency on sys/queue.h on freebsd (patch by Vanilla Hsu).
277 - do async init within ev_async_start, not ev_async_set, which avoids
278 an API quirk where the set function must be called in the C++ API
279 even when there is nothing to set.
280 - add (undocumented) EV_ENABLE when adding events with kqueue,
281 this might help with OS X, which seems to need it despite documenting
282 not to need it (helpfully pointed out by Tilghman Lesher).
283 - do not use poll by default on freebsd, it's broken (what isn't
284 on freebsd...).
285 - allow to embed epoll on kernels >= 2.6.32.
286 - configure now prepends -O3, not appends it, so one can still
287 override it.
288 - ev.pod: greatly expanded the portability section, added a porting
289 section, a description of watcher states and made lots of minor fixes.
290 - disable poll backend on AIX, the poll header spams the namespace
291 and it's not worth working around dead platforms (reported
292 and analyzed by Aivars Kalvans).
293 - improve header file compatibility of the standalone eventfd code
294 in an obscure case.
295 - implement EV_AVOID_STDIO option.
296 - do not use sscanf to parse linux version number (smaller, faster,
297 no sscanf dependency).
298 - new EV_CHILD_ENABLE and EV_SIGNAL_ENABLE configurable settings.
299 - update libev.m4 HAVE_CLOCK_SYSCALL test for newer glibcs.
300 - add section on accept() problems to the manpage.
301 - rename EV_TIMEOUT to EV_TIMER.
302 - rename ev_loop_count/depth/verify/loop/unloop.
303 - remove ev_default_destroy and ev_default_fork.
304 - switch to two-digit minor version.
305 - work around an apparent gentoo compiler bug.
306 - define _DARWIN_UNLIMITED_SELECT. just so.
307 - use enum instead of #define for most constants.
308 - improve compatibility to older C++ compilers.
309 - (experimental) ev_run/ev_default_loop/ev_break/ev_loop_new have now
310 default arguments when compiled as C++.
311 - enable automake dependency tracking.
312 - ev_loop_new no longer leaks memory when loop creation failed.
313 - new ev_cleanup watcher type.
314
3153.9 Thu Dec 31 07:59:59 CET 2009
316 - signalfd is no longer used by default and has to be requested
317 explicitly - this means that easy to catch bugs become hard to
318 catch race conditions, but the users have spoken.
319 - point out the unspecified signal mask in the documentation, and
320 that this is a race condition regardless of EV_SIGNALFD.
321 - backport inotify code to C89.
322 - inotify file descriptors could leak into child processes.
323 - ev_stat watchers could keep an erroneous extra ref on the loop,
324 preventing exit when unregistering all watchers (testcases
325 provided by ry@tinyclouds.org).
326 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration
327 symbols to make it easier for apps to do their own fd management.
328 - support EV_IDLE_ENABLE being disabled in ev++.h
329 (patch by Didier Spezia).
330 - take advantage of inotify_init1, if available, to set cloexec/nonblock
331 on fd creation, to avoid races.
332 - the signal handling pipe wasn't always initialised under windows
333 (analysed by lekma).
334 - changed minimum glibc requirement from glibc 2.9 to 2.7, for
335 signalfd.
336 - add missing string.h include (Denis F. Latypoff).
337 - only replace ev_stat.prev when we detect an actual difference,
338 so prev is (almost) always different to attr. this might
339 have caused the problems with 04_stat.t.
340 - add ev::timer->remaining () method to C++ API.
341
3423.8 Sun Aug 9 14:30:45 CEST 2009
343 - incompatible change: do not necessarily reset signal handler
344 to SIG_DFL when a sighandler is stopped.
345 - ev_default_destroy did not properly free or zero some members,
346 potentially causing crashes and memory corruption on repeated
347 ev_default_destroy/ev_default_loop calls.
348 - take advantage of signalfd on GNU/Linux systems.
349 - document that the signal mask might be in an unspecified
350 state when using libev's signal handling.
351 - take advantage of some GNU/Linux calls to set cloexec/nonblock
352 on fd creation, to avoid race conditions.
353
3543.7 Fri Jul 17 16:36:32 CEST 2009
355 - ev_unloop and ev_loop wrongly used a global variable to exit loops,
356 instead of using a per-loop variable (bug caught by accident...).
357 - the ev_set_io_collect_interval interpretation has changed.
358 - add new functionality: ev_set_userdata, ev_userdata,
359 ev_set_invoke_pending_cb, ev_set_loop_release_cb,
360 ev_invoke_pending, ev_pending_count, together with a long example
361 about thread locking.
362 - add ev_timer_remaining (as requested by Denis F. Latypoff).
363 - add ev_loop_depth.
364 - calling ev_unloop in fork/prepare watchers will no longer poll
365 for new events.
366 - Denis F. Latypoff corrected many typos in example code snippets.
367 - honor autoconf detection of EV_USE_CLOCK_SYSCALL, also double-
368 check that the syscall number is available before trying to
369 use it (reported by ry@tinyclouds).
370 - use GetSystemTimeAsFileTime instead of _timeb on windows, for
371 slightly higher accuracy.
372 - properly declare ev_loop_verify and ev_now_update even when
373 !EV_MULTIPLICITY.
374 - do not compile in any priority code when EV_MAXPRI == EV_MINPRI.
375 - support EV_MINIMAL==2 for a reduced API.
376 - actually 0-initialise struct sigaction when installing signals.
377 - add section on hibernate and stopped processes to ev_timer docs.
378
3793.6 Tue Apr 28 02:49:30 CEST 2009
380 - multiple timers becoming ready within an event loop iteration
381 will be invoked in the "correct" order now.
382 - do not leave the event loop early just because we have no active
383 watchers, fixing a problem when embedding a kqueue loop
384 that has active kernel events but no registered watchers
385 (reported by blacksand blacksand).
386 - correctly zero the idx values for arrays, so destroying and
387 reinitialising the default loop actually works (patch by
388 Malek Hadj-Ali).
389 - implement ev_suspend and ev_resume.
390 - new EV_CUSTOM revents flag for use by applications.
391 - add documentation section about priorities.
392 - add a glossary to the documentation.
393 - extend the ev_fork description slightly.
394 - optimize a jump out of call_pending.
395
3963.53 Sun Feb 15 02:38:20 CET 2009
397 - fix a bug in event pipe creation on win32 that would cause a
398 failed assertion on event loop creation (patch by Malek Hadj-Ali).
399 - probe for CLOCK_REALTIME support at runtime as well and fall
400 back to gettimeofday if there is an error, to support older
401 operating systems with newer header files/libraries.
402 - prefer gettimeofday over clock_gettime with USE_CLOCK_SYSCALL
403 (default most everywhere), otherwise not.
2 404
33.52 Wed Jan 7 21:43:02 CET 2009 4053.52 Wed Jan 7 21:43:02 CET 2009
4 - fix compilation of select backend in fd_set mode when NFDBITS is 406 - fix compilation of select backend in fd_set mode when NFDBITS is
5 missing (to get it to compile on QNX, reported by Rodrigo Campos). 407 missing (to get it to compile on QNX, reported by Rodrigo Campos).
6 - better select-nfds handling when select backend is in fd_set mode. 408 - better select-nfds handling when select backend is in fd_set mode.
… …
11 - actually verify that local and remote port are matching in 413 - actually verify that local and remote port are matching in
12 libev's socketpair emulation, which makes denial-of-service 414 libev's socketpair emulation, which makes denial-of-service
13 attacks harder (but not impossible - it's windows). Make sure 415 attacks harder (but not impossible - it's windows). Make sure
14 it even works under vista, which thinks that getpeer/sockname 416 it even works under vista, which thinks that getpeer/sockname
15 should return fantasy port numbers. 417 should return fantasy port numbers.
16 - include "libev" all assertion messages for potentially 418 - include "libev" in all assertion messages for potentially
17 clearer diagnostics. 419 clearer diagnostics.
18 - event_get_version (libevent compatibility) returned 420 - event_get_version (libevent compatibility) returned
19 a useless string instead of the expected version string 421 a useless string instead of the expected version string
20 (patch by W.C.A. Wijngaards). 422 (patch by W.C.A. Wijngaards).
21 423
… …
71 do similar changes for event.c. 473 do similar changes for event.c.
72 474
733.45 Tue Oct 21 21:59:26 CEST 2008 4753.45 Tue Oct 21 21:59:26 CEST 2008
74 - disable inotify usage on linux <2.6.25, as it is broken 476 - disable inotify usage on linux <2.6.25, as it is broken
75 (reported by Yoann Vandoorselaere). 477 (reported by Yoann Vandoorselaere).
76 - ev_stat errornously would try to add inotify watchers 478 - ev_stat erroneously would try to add inotify watchers
77 even when inotify wasn't available (this should only 479 even when inotify wasn't available (this should only
78 have a performance impact). 480 have a performance impact).
79 - ev_once now passes both timeout and io to the callback if both 481 - ev_once now passes both timeout and io to the callback if both
80 occur concurrently, instead of giving timeouts precedence. 482 occur concurrently, instead of giving timeouts precedence.
81 - disable EV_USE_INOTIFY when sys/inotify.h is too old. 483 - disable EV_USE_INOTIFY when sys/inotify.h is too old.
… …
149 - enable inlining with gcc versions 2 and 3. 551 - enable inlining with gcc versions 2 and 3.
150 - work around broken poll implementations potentially 552 - work around broken poll implementations potentially
151 not clearing revents field in ev_poll (Brandon Black) 553 not clearing revents field in ev_poll (Brandon Black)
152 (no such systems are known at this time). 554 (no such systems are known at this time).
153 - work around a bug in realloc on openbsd and darwin, 555 - work around a bug in realloc on openbsd and darwin,
154 also makes the errornous valgrind complaints 556 also makes the erroneous valgrind complaints
155 go away (noted by various people). 557 go away (noted by various people).
156 - fix ev_async_pending, add c++ wrapper for ev_async 558 - fix ev_async_pending, add c++ wrapper for ev_async
157 (based on patch sent by Johannes Deisenhofer. 559 (based on patch sent by Johannes Deisenhofer).
158 - add sensible set method to ev::embed. 560 - add sensible set method to ev::embed.
159 - made integer constants type int in ev.h. 561 - made integer constants type int in ev.h.
160 562
1613.2 Wed Apr 2 17:11:19 CEST 2008 5633.2 Wed Apr 2 17:11:19 CEST 2008
162 - fix a 64 bit overflow issue in the select backend, 564 - fix a 64 bit overflow issue in the select backend,

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines