| 1 | Revision history for Perl extension Coro. |
1 | Revision history for Perl extension Coro. |
| 2 | |
2 | |
| 3 | TODO: should explore PerlIO::via::CoroCede and PerlIO::via::CoroAIO. |
3 | TODO: should explore PerlIO::coroaio (perl leaks like hell). |
| 4 | TODO: maybe implement a default message channel, very much like Erlang's |
4 | TODO: maybe implement a default message channel, very much like Erlang's |
| 5 | actor model (which is cool in a lot of important aspects (failures!), |
5 | actor model (which is cool in a lot of important aspects (failures!), |
| 6 | but very lacking in others (higher level ipc)). |
6 | but very lacking in others (higher level ipc)). |
|
|
7 | TODO: unready_all |
|
|
8 | TODO: myhttpd header parsing |
|
|
9 | TODO: channel->maxsize(newsize)? |
| 7 | |
10 | |
|
|
11 | 5.2 |
|
|
12 | - Coro::Storable destroyed the prototypes of the functions it wrapped. |
|
|
13 | - export rouse_cb and rouse_wait by default now. |
|
|
14 | - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle. |
|
|
15 | - new method $state->swap_sv. |
|
|
16 | - added section on "windows process emulation" to the manpage, |
|
|
17 | after a not-so-fruitful (nor-friendly) "discussion" with chip |
|
|
18 | salzenberg (discussion implies arguments, but his only arguments |
|
|
19 | were ad-hominems, one wonders why he started it in the first |
|
|
20 | place). I hope this explains it well enough for him to understand, |
|
|
21 | and maybe well enough for others to understand. |
|
|
22 | - use common::sense everywhere now. |
|
|
23 | - idle callbacks are no longer supported, use idle coros instead. |
|
|
24 | - print a thread listing when a deadlock is detected. |
|
|
25 | |
|
|
26 | 5.17 Sat Aug 22 23:09:31 CEST 2009 |
|
|
27 | - work around a bug in the perl debugger causing crashes |
|
|
28 | when running under the debugger by marking _pool_handler as nodebug. |
|
|
29 | - speed up Coro::async considerably. |
|
|
30 | - try some hacks to get netbsd to work "more often" - their broken |
|
|
31 | setjmp/longjmp, ucontext *and* phtreads are really hard on Coro. |
|
|
32 | - convert Coro to AE 5.0 API. |
|
|
33 | |
|
|
34 | 5.162 Tue Jul 28 04:04:03 CEST 2009 |
|
|
35 | - perl 5.8.2 is now minimum requirement. |
|
|
36 | - skip t/19_handle.t on broken windows perls. |
|
|
37 | |
|
|
38 | 5.161 Wed Jul 22 04:47:38 CEST 2009 |
|
|
39 | - Coro::AnyEvent::poll could have a different prototype when EV was |
|
|
40 | used as backend (analyzed by Tatsuhiko Miyagawa). |
|
|
41 | - Coro::AnyEvent errornously initialised the event loop when loaded, |
|
|
42 | not on demand. |
|
|
43 | - try to workaround rare */t/01_unblock.t failures. |
|
|
44 | |
|
|
45 | 5.16 Tue Jul 21 01:44:37 CEST 2009 |
|
|
46 | - Coro::AnyEvent failed to hook into the event loop |
|
|
47 | when no threads had been readied between detecting |
|
|
48 | the event loop and actually running it. |
|
|
49 | - considerably speed up Coro::Select by taking avdantage |
|
|
50 | of AnyEvent > 4.8 and some other optimisations. |
|
|
51 | - implement paragraph readline mode in Coro::Handle |
|
|
52 | (based on patches by Zsbán Ambrus). |
|
|
53 | - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported |
|
|
54 | and analyzed by Yasuhiro MATSUMOTO). |
|
|
55 | - clarified libcoro license and copyright. |
|
|
56 | - someone stole my EXTRA_META!!! |
|
|
57 | - implement Coro::Select::patch_pp_sselect and it's brother, |
|
|
58 | for hardcode select overriding. |
|
|
59 | |
|
|
60 | 5.151 Mon Jul 6 05:41:57 CEST 2009 |
|
|
61 | - backport to windows process emulation code again (patch by |
|
|
62 | Yasuhiro MATSUMOTO). |
|
|
63 | - slightly update Coro::MakeMaker. |
|
|
64 | |
|
|
65 | 5.15 Tue Jun 30 10:28:06 CEST 2009 |
|
|
66 | - deprecate Coro::Socket, document how to get ipv6 support via |
|
|
67 | AnyEvent::Socket instead. |
|
|
68 | - implement signal->wait ($cb) interface, similar to semaphores. |
|
|
69 | - work around SvOK not supporting getmagic, so we have to getmagic |
|
|
70 | to test for undef :/ (reported by Matthias Waldorf). |
|
|
71 | - load Coro::AnyEvent in all modules using AnyEvent. |
|
|
72 | - work around perl corrupting our internal data structures, |
|
|
73 | reported by Tokuhiro Matsuno. |
|
|
74 | - enable per-coroutine real and cpu time gathering |
|
|
75 | (Coro::State::enable_times). |
|
|
76 | |
|
|
77 | 5.14 Wed Jun 24 01:37:48 CEST 2009 |
|
|
78 | - provide explicit functions to "cede" to the event loop in Coro::AnyEvent, |
|
|
79 | as this seems to have been a difficult concept (poll, sleep, idle, |
|
|
80 | idle_upto). |
|
|
81 | - add Coro::AnyEvent::readable/writable functions. |
|
|
82 | - clarify Coro::EV/Event/AnyEvent manpages. |
|
|
83 | - free per-thread global scalars in the thread calling ->cancel, to |
|
|
84 | avoid crashes when $_, $@ etc., are magical but some of those |
|
|
85 | had already been freed. |
|
|
86 | - "unexperimentalise" the callback interface for Coro::Semaphore. |
|
|
87 | - speed up ready queue management/context switching by using a linked |
|
|
88 | list instead of an array (~5-10%). |
|
|
89 | - implement "watch" command in Coro::Debug shells. |
|
|
90 | - for fun, implement time-slicing as an example in the manpage. |
|
|
91 | - if AnyEvent detects EV or Event, but we don't have Coro::EV or |
|
|
92 | Coro::Event, use the normal AnyEvent handling instead of dieing |
|
|
93 | (the same is true for Coro::Handle). |
|
|
94 | - properly document Coro::EV::timed_io_once. |
|
|
95 | - avoid unneccessary ->cancel calls in Coro::Handle. |
|
|
96 | - maybe make it work on mingw32 with win32 backend |
|
|
97 | (based on patch by Yasuhiro Matsumoto). |
|
|
98 | |
|
|
99 | 5.132 Fri May 29 09:00:39 CEST 2009 |
|
|
100 | - do not keep a reference to the argument itself in |
|
|
101 | Coro::Semaphore::guard, as it could change later. |
|
|
102 | - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options, |
|
|
103 | should support a prepare callback. |
|
|
104 | |
|
|
105 | 5.131 Mon Mar 16 23:20:37 CET 2009 |
|
|
106 | - implement and document Coro->suspend, Coro->resume. |
|
|
107 | - fix Coro::Select implementation to not (often) close |
|
|
108 | the passed file descriptors (testcase provided by pippijn). |
|
|
109 | |
|
|
110 | 5.13 Mon Dec 15 21:51:42 CET 2008 |
|
|
111 | - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave). |
|
|
112 | - don't set diehook to C<undef> but instead to NULL, to avoid |
|
|
113 | spurious warnings. |
|
|
114 | - fix a lot of bugs in Coro::SemaphoreSet. |
|
|
115 | - Coro::SemaphoreSet will less often create a semaphore needlessly. |
|
|
116 | - add Coro::SemaphoreSet::count and wait methods. |
|
|
117 | - take advantage of the new Guard module. |
|
|
118 | - deprecate Coro::guard. |
|
|
119 | - try to fix the dreaded 01_unblock tests once more. I hate it when |
|
|
120 | testsuites need more fixing than the code they are supposed to test. |
|
|
121 | - croak in more cases when a required callback isn't resolvable. |
|
|
122 | - fix some minor issues in Coro::State->call/eval. |
|
|
123 | - use current coroutine context instead of a temporary one |
|
|
124 | when temporarily switching to another coroutine. |
|
|
125 | - do not call C-level on_destroy handlers during global destruction, |
|
|
126 | to avoid needless segfaults. |
|
|
127 | |
|
|
128 | 5.12 Sun Dec 7 13:30:38 CET 2008 |
|
|
129 | - add default config for MirOS, which seems to be a bug-to-bug |
|
|
130 | compatible fork of openbsd ("world domination by releasing |
|
|
131 | openbsd cvs before the openbsd folks do it" or so :). |
|
|
132 | - free_padlist did destroy the names pad, not good, but didn't |
|
|
133 | seem to bother perl - this could fix issues such as eval "" |
|
|
134 | inside a function called from multiple coroutines. |
|
|
135 | - use a different method to detect destruction time. |
|
|
136 | - be more careful when freeing padlists just before global |
|
|
137 | destruction. |
|
|
138 | - fixed and expanded the call/cc example. |
|
|
139 | - renamed _terminate to _coro_run. |
|
|
140 | - new method Coro::Channel->shutdown. |
|
|
141 | - try pthreads on openbsd <4.4 (broken sigaltstack, will |
|
|
142 | pthreads fare better?). |
|
|
143 | - be less picky about destroying "a" running coroutine. |
|
|
144 | |
|
|
145 | 5.11 Tue Nov 25 21:49:05 CET 2008 |
|
|
146 | - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly |
|
|
147 | working code with failed assertions, introducing more bugs than |
|
|
148 | it fixes, requiring elaborate workarounds :( |
|
|
149 | |
|
|
150 | 5.1 Mon Nov 24 08:54:59 CET 2008 |
|
|
151 | - wrote a small introductory tutorial - Coro::Intro. |
|
|
152 | - convert Coro::Timer, Coro::Select and Coro::Util to rouse API. |
|
|
153 | - Coro::Select did errornously dup the file descriptors |
|
|
154 | and didn't work with all AnyEvent backends. |
|
|
155 | - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking |
|
|
156 | LWP data transfers. |
|
|
157 | - disassociate c contexts from coro objects - this is agruably more |
|
|
158 | correct, but mostly allows sharing of cctxs between coro and state |
|
|
159 | objects, for added memory savings and speed increases. |
|
|
160 | - bumped $Coro::POOL_RSS up to 32kb by default. |
|
|
161 | - no longer set the optype to OP_CUSTOM, as B::* understandably |
|
|
162 | doesn't like this very much (and we *are* a type of entersub). |
|
|
163 | - implement state cloning, just to prove that call/cc can be done. |
|
|
164 | - automatically load Coro::AnyEvent in Coro::Handle. |
|
|
165 | - wrap ->cancel calls in eval inside Coro::Handle as EV watchers |
|
|
166 | do not have this method (and don't need it either). |
|
|
167 | - speed up generic anyevent methods in Coro::Handle by using rouse |
|
|
168 | callbacks. |
|
|
169 | - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and |
|
|
170 | others. It also makes the debugger happier, as you can trace |
|
|
171 | through the idle threads now. |
|
|
172 | - add comppad_name* and hints ($^H, %^H) to per-thread variables. |
|
|
173 | - eg/event was pretty broken. |
|
|
174 | - better 5.8.6 compatibility. |
|
|
175 | |
|
|
176 | 5.0 Thu Nov 20 10:35:05 CET 2008 |
|
|
177 | - NEW ARCHITECTURE: use the latest 4.x version if you experience |
|
|
178 | stability issues. |
|
|
179 | - bump API version to 7 - all dependents must be recompiled. |
|
|
180 | - removed timed_* functions - they were not being used anyways |
|
|
181 | and should be replaced by a more generic mechanism - |
|
|
182 | and were annoying to support anyways :) |
|
|
183 | - removed SemaphoreSet's waiter method - use sem method instead. |
|
|
184 | - Coro::Semaphore->adjust didn't correctly wake up enough waiters. |
|
|
185 | - async_pool did free a scalar value twice |
|
|
186 | ("Attempt to unreference..."). |
|
|
187 | - fix a longstanding bug where calling terminate on a coro that |
|
|
188 | was waiting for a semaphore that was just becoming available |
|
|
189 | would cause a deadlock (semaphore would get into a state where |
|
|
190 | it was available but waiters were still blocked). |
|
|
191 | - calling throw on a coroutine that is waiting for a semaphore will |
|
|
192 | no longer make it acquire the semaphore (and thus leak a count). |
|
|
193 | - perl's process emulation is now not even theoretically supported |
|
|
194 | anymore. |
|
|
195 | - new functions Coro::rouse_cb and Coro::rouse_wait for easier |
|
|
196 | conversion of callback-style to blocking-style. |
|
|
197 | - new methods $coro->schedule_to and ->cede_to, to specifically |
|
|
198 | schedule or cede to a specific coroutine. |
|
|
199 | - new function Coro::Semaphore::wait. |
|
|
200 | - use named constants in Coro::Channel (Richard Hundt). |
|
|
201 | - directly patch the entersub opcode calling SLF functions (cede, |
|
|
202 | transfer and so on). this does speed up context switching, but |
|
|
203 | more importanly, it frees us from the hardcoded behaviour of |
|
|
204 | entersub, so we might actually be able to return something from |
|
|
205 | those functions and atcually create new ones. |
|
|
206 | - take advantage of __builtin_frame_address on gcc. |
|
|
207 | - expose THX in coroapi (not sure whether this was a wise decision, |
|
|
208 | as "threaded" perls are running at half speed anyways). |
|
|
209 | - implement execute_slf (schedule-like-function) interface that makes |
|
|
210 | it possible to implement schedule-like-functions in XS. |
|
|
211 | - use new SLF interface to massively speed up Coro::EV by roughly a |
|
|
212 | factor of two. |
|
|
213 | - used new SLF interface to massively speed up Coro::Semaphore by a |
|
|
214 | factor of three. |
|
|
215 | - used new SLF interface to speed up Coro::AIO by roughly a factor of |
|
|
216 | four and reduce its memory usage considerably. |
|
|
217 | - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore, |
|
|
218 | for a nice speedup and vastly more correct behaviour. Also implement |
|
|
219 | a new method "sem" to get at the underlying semaphore object. |
|
|
220 | - implement Coro::Channel in terms of Coro::Semaphore, for a moderate |
|
|
221 | (in comparison) 20-40% speedup. |
|
|
222 | - used new SLF interface to reimplement Coro::Signal gaining |
|
|
223 | some unknown (because I was too lazy), but certain, speedup, and also |
|
|
224 | making signals reliable for the first time. |
|
|
225 | - used new SLF interface and other optimisations to speed up async_pool |
|
|
226 | by a factor of two. It also doesn't rely on perl's exception mechanism |
|
|
227 | to exit anymore. The overhead for terminating an async_pool, coro over |
|
|
228 | a normal async is now very small. |
|
|
229 | - sped up coroutine creation/destruction by 40%. |
|
|
230 | - forgot to include Coro/libcoro/README in the dist for all these years. |
|
|
231 | - work around a freebsd pthreads bug (manual testcancel is required as |
|
|
232 | pthread_cond_wait isn't a cancellation point on freebsd). |
|
|
233 | - use new rouse functions to speed up and simplify Coro::BDB. |
|
|
234 | - make "prefer perl native functions" work with threaded perls. |
|
|
235 | - condense Coro::Debug ps output, hint at v and w flags. |
|
|
236 | - (libcoro) lots of minor cleanups and portability improvements. |
|
|
237 | |
|
|
238 | 4.914 Wed Nov 19 12:54:18 CET 2008 |
|
|
239 | - fix a disastrous bug in the readline optimisation |
|
|
240 | introduced in 4.801. |
|
|
241 | |
|
|
242 | 4.913 Sat Nov 15 07:58:28 CET 2008 |
|
|
243 | - async_pool did free a scalar value twice |
|
|
244 | ("Attempt to unreference..."). |
|
|
245 | |
|
|
246 | 4.912 Thu Nov 13 18:31:23 CET 2008 |
|
|
247 | - minor cleanups. |
|
|
248 | - use much larger stacks on linux and perl < 5.8.8. |
|
|
249 | - Coro::Debug::new_unix_server did not unlink the socket |
|
|
250 | when destroyed. |
|
|
251 | |
|
|
252 | 4.911 Tue Nov 11 04:26:17 CET 2008 |
|
|
253 | - "port" to threaded perls. |
|
|
254 | |
|
|
255 | 4.91 Mon Nov 10 05:36:38 CET 2008 |
|
|
256 | - the ->throw exception object no longer leaks. |
|
|
257 | - creating a new cctx leaked a scopestack entry (memleak). |
|
|
258 | - new coroutines didn't get created with a zero flags field |
|
|
259 | (unknown impact). |
|
|
260 | - calling ->throw on a not-yet-started coroutine should now work |
|
|
261 | instead of being ignored. |
|
|
262 | - ->throw is now supported on Coro::State objects. |
|
|
263 | - clean up cctx creation code a bit. |
|
|
264 | - entersub is actually a UNOP, not a LOGOP (not a bugfix). |
|
|
265 | |
|
|
266 | 4.9 Sat Nov 8 17:45:27 CET 2008 |
|
|
267 | - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). |
|
|
268 | - (libcoro) no longer leak threads in the experimental pthread backend, |
|
|
269 | also speed it up considerably. |
|
|
270 | - (libcoro) do not rely on makecontext passing void *'s unscathed. |
|
|
271 | - fix compiletime dependencies on libcoro in the Makefile. |
|
|
272 | - cctx_count wasn't always updated properly. |
|
|
273 | - Coro::State::cctx_stacksize wasn't applied correctly. |
|
|
274 | - new function Coro::State::cctx_max_idle. |
|
|
275 | - the default max number of idle C contexts is now 4. |
|
|
276 | - (libcoro) try harder to get _setjmp/_longjmp. |
|
|
277 | - (libcoro) cleanup and extend the libcoro API to officially |
|
|
278 | allow the creation of empty source contexts. |
|
|
279 | - very experimental workaround for the totally broken netbsd platform. |
|
|
280 | - tried to hack around openbsd bugs. |
|
|
281 | |
|
|
282 | 4.804 Wed Nov 5 16:36:00 CET 2008 |
|
|
283 | - Coro::Debug::new_unix_server would not create a non-blocking listening |
|
|
284 | socket, sometimes causing freezes. |
|
|
285 | - (libcoro) fix misaligned stack points for setjmp and assembly |
|
|
286 | methods, which can cause crashes on x86/x86_64 with a sufficiently |
|
|
287 | aggressive compiler. |
|
|
288 | - new function: Coro::Debug::new_tcp_server. |
|
|
289 | - move ->throw into the Coro class because it only works on coro objects. |
|
|
290 | |
|
|
291 | 4.803 Mon Nov 3 17:16:12 CET 2008 |
|
|
292 | - (libcoro) use a global asm statement to become independent of gcc |
|
|
293 | otpimisations for CORO_ASM (thanks to pippijn for the idea). |
|
|
294 | - try to workaround yet another broken bsd, this time dragonfly. |
|
|
295 | |
|
|
296 | 4.802 Thu Oct 30 10:56:12 CET 2008 |
|
|
297 | - support -fno-omit-frame-pointer on x86 with the assembly method. |
|
|
298 | - tune 01_unblock.t tests a bit. |
|
|
299 | |
|
|
300 | 4.801 Wed Oct 22 18:33:37 CEST 2008 |
|
|
301 | - improve readline speed for very long "lines". |
|
|
302 | - backport to 5.8.8. |
|
|
303 | |
|
|
304 | 4.8 Thu Oct 2 13:34:40 CEST 2008 |
|
|
305 | - new function Coro::AIO::aio_wait. |
|
|
306 | - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent". |
|
|
307 | - greatly speed up and reduce memory usage of Coro::AIO requests. |
|
|
308 | - implement some other µ-optimisations. |
|
|
309 | |
|
|
310 | 4.749 Mon Sep 29 14:40:12 CEST 2008 |
|
|
311 | - port to slow and broken pseudo-threaded perls. (courtesy pippijn). |
|
|
312 | |
|
|
313 | 4.748 Sat Sep 27 14:03:19 CEST 2008 |
|
|
314 | - implement, but do not document, PerlIO::cede(granularity). |
|
|
315 | - Coro::Storable forgot to wrap Storable::pstore. |
|
|
316 | - work around the multitude of leaks and memory corruption |
|
|
317 | bugs in PerlIO::via by using our own C-level perliol. As a side |
|
|
318 | effect, Coro::Storable is now much, much, much faster. |
|
|
319 | |
|
|
320 | 4.747 Tue Sep 23 01:59:41 CEST 2008 |
|
|
321 | - fix a per-cv memleak (one empty array was leaked per |
|
|
322 | code reference). |
|
|
323 | - avoid a crash in coro->call|rss when the coroutine was already |
|
|
324 | destroyed (most noticably when using Coro::Debug::ps :) |
|
|
325 | - also protect *Storable::FILE. |
|
|
326 | - push up default storable granularity to 20ms. |
|
|
327 | |
|
|
328 | 4.746 Sun Sep 21 03:22:20 CEST 2008 |
|
|
329 | - be more insistent on locking Storable against reentrancy |
|
|
330 | in Coro::Storable. |
|
|
331 | - move swap_def?v and throw to Coro::State, as documented. |
|
|
332 | |
|
|
333 | 4.745 Thu Jul 24 00:14:38 CEST 2008 |
|
|
334 | - remove debugging code related to MgPV_nolen_const, also try to |
|
|
335 | make it compile with perl 5.8.6 (yes, apple apparently loves |
|
|
336 | outdated software). Reported by John S. |
|
|
337 | |
|
|
338 | 4.744 Tue Jul 8 22:06:35 CEST 2008 |
|
|
339 | - correctly provide default DIE/WARN handlers as documented. |
|
|
340 | - also overwrite PL_vtbl_sigelem.svt_clear, even though current |
|
|
341 | implementations inside perl work fine for us. |
|
|
342 | |
|
|
343 | 4.743 Mon Jun 16 00:21:57 CEST 2008 |
|
|
344 | - when using Coro::EV without running EV::loop it could |
|
|
345 | result in busy-waiting for events, this has been fixed. |
|
|
346 | - reduce codesize and improve performance by using EV_DEFAULT_UC. |
|
|
347 | |
|
|
348 | 4.742 Sat May 31 14:10:21 CEST 2008 |
|
|
349 | - implement a workaround for (some) perl <5.8.8 versions. |
|
|
350 | - require EV 3.3+. |
|
|
351 | |
|
|
352 | 4.741 Fri May 30 23:33:09 CEST 2008 |
|
|
353 | - tell netbsd how utterly broken their imitation of an OS is |
|
|
354 | and refuse to build by default if pthreads are in use. |
|
|
355 | - switch to "s" method on all bsd's by default, as their ucontext |
|
|
356 | stuff seems just too broken. |
|
|
357 | - fix a bug in Coro::Select. |
|
|
358 | |
|
|
359 | 4.74 Thu May 29 20:05:31 CEST 2008 |
|
|
360 | - do not test Coro::LWP for lack of dependencies. |
|
|
361 | - do not test Coro::Select for lack of working perls. |
|
|
362 | |
|
|
363 | 4.73 Thu May 29 2008 |
|
|
364 | - fix a bug in Coro::EV which would cause it to block despite |
|
|
365 | there being runnable coroutines. |
| 8 | - sprinkle "no warnings" freely over everything, also suppress |
366 | - sprinkle "no warnings" freely over everything, also suppress |
| 9 | warnings for some other modules. |
367 | warnings for some other modules. |
|
|
368 | - fix typo in WSAEWOULDBLOCK. |
| 10 | |
369 | |
| 11 | 4.72 Sun May 25 05:14:36 CEST 2008 |
370 | 4.72 Sun May 25 05:14:36 CEST 2008 |
| 12 | - tweak META.yaml a bit, unfortunately, there is no documented way |
371 | - tweak META.yaml a bit, unfortunately, there is no documented way |
| 13 | to have optional dependencies with CPAN. doh :( |
372 | to have optional dependencies with CPAN. doh :( |
| 14 | - avoid running some tests on windows because they would fail due to |
373 | - avoid running some tests on windows because they would fail due to |
