1 Revision history for the Perl extension Time::HiRes.
4 - in t/clock.t in a fast system we need to burn more CPU,
5 reported and fix suggested by Joel C. Maslak <jmaslak@antelope.net>
7 - for t/utime.t in netbsd we need /sbin in PATH to find the mount command,
8 this is needed for the fix in 1.9751 to actually help netbsd with noatime
10 (thanks to Nigel Horne)
13 - fix an error in the error message of utimensat() not available:
14 it said futimens() not available
15 - add --force alias for Makefile.PL --configure
18 - in macOS/OSX/Darwin, use __has_builtin() check also for utimensat(),
20 "HiRes.xs:1474:16: error: unrecognized platform name macOS"
22 (oversight from 1.9749)
23 - do not define TIME_HIRES_STAT/d_hires_stat if none was found, instead
24 of defining it to be zero, which case has no implementation in hrstatns()
25 (thanks to Nigel Horne)
26 - in t/utime.t try to divine if the filesystem of the tempfiles has been
27 mounted with the 'noatime' option, which can prohibit updating the
28 access time timestamp. Also document this in HiRes.pm.
29 (thanks to Nigel Horne, original analysis by Slaven Rezic)
30 - synchronize the constant lists in HiRes.pm:@EXPORT_OK
31 and Makefile.PL:doConstants and regenerate fallback/const-c.inc
32 and fallback/const-xs.inc, this fixes Perl 5.6.2 issue with
33 d_futimens not allegedly being a valid macro in t/utime.t
34 (using Perl 5.26.1 for the regenerating, not 5.6.2)
35 (thanks to Nigel Horne)
36 - in t/utime.t define a nop sub done_testing for ancient Perls
38 - in Perl 5.6.2 a bogus warning
39 "Use of uninitialized value in subroutine entry"
40 is issued from t/alarm.t: add a comment documenting that
43 - update META.yml and META.json for XSLoader instead of DynaLoader
48 - Quieten Time-HiRes test noise: blead 0f009486
49 - Switch some core modules to XSLoader: blead b9a5a78f [rt.cpan.org #129789]
50 - disable a flapping test for CI: blead cd2a978d
51 - Avoid newGVgen in blead-upstream modules: blead 732d3893
52 - Declaration after statement in typemap: blead d269f586
53 - semicolon-friendly diagnostic control: blead 7347ee54
54 - in macOS (OS X) check properly in runtime whether futimens()
55 and utimensat() are available
59 - semicolon-friendly diagnostic control: blead 7347ee54
63 - Switch some core modules to XSLoader: blead b9a5a78f [rt.cpan.org #129789]
66 - Unreliable t/usleep.t and t/utime.t tests [rt.cpan.org #122819]
67 Avoid testing for $dt = $t2 - $t1 and assuming $dt is less than
68 some value since a heavily loaded machine can delay the $t2.
69 Testing for greater than is fine.
70 - Tweak the configuring messages about subsecond stat/utime.
73 - Properly check for futimens/utimensat actually doing something.
74 This should handle 'gnukfreebsd' (which has only stubs, so far
75 we have been skipping the test) and as a new thing 'gnu' (Hurd)
76 (also only stubs). Thanks to Nigel Horne.
77 - Scan in t/utime.t whether the filesystem (wherever File::Temp
78 puts it tempfiles) supports subsecond timestamps. This removes
79 the fragile Linux/ext3 specific hack. As a side effect, the
80 setting of subsecond timestamps is tested only if reading of
81 them is supported. Thanks to Carsten Gaebler for the test idea,
82 and Ryan Voots for testing.
85 - add more potential clock constants, like CLOCK_MONOTONIC_FAST
86 (available in FreeBSD), and not all potentially found clock
87 constants were properly exported to be available from Perl,
88 see your system's clock_gettime() documentation for the available ones
91 - correct declared minimum Perl version (should be 5.6, was declared
92 as 5.8 since 1.9727_03): blead af94b3ac
93 - fix the fix for 'do file' to load hints in Makefile.PL: blead 3172fdbc
96 - prefer 3-argument open: blead 1ae6ead9
97 - fix dist/Time-HiRes/t/*.t that assumed '.' in @INC: blead 465db51d
98 - fix cases where 'do file' should be 'do ./file'.: blead 8b69401c
101 - C++11 compatibility: blead a914236c
102 - El Capitan compatibility: blead 45bbc013
103 - use CLOCK_REALTIME for clock_nanosleep scan: blead 56ed632c
104 - include file consistency in scans: blead 605b4dbe
105 - use clockid_t consistently: blead 2979fdf3
106 - use hv_fetchs(): blead 07e4dd7a
107 - scan for clockid_t (needed for macos Sierra): blead 69d1f2c2
108 - darwin lacks clockid_t: blead d1e7b5aa
109 [rt.cpan.org #129789]
110 - provide template for tempfile: blead 91dca83e
111 - explicit cast for clockid_t in C++11: blead 1c566547
112 - upgrade to a newer ppport (3.35)
115 - the ext3/ext2 filesystems do not have subsecond resolution,
116 therefore skip the t/utime.t test
117 [rt.cpan.org #116127]
120 - the upcoming macOS 10.12 (Sierra, the operating system formerly
121 known as OS X, or Darwin) has implemented the clock_gettime()
122 and clock_getres() interfaces but not the clock_nanosleep()
125 - fix the dates in the Changes, it's not July yet
126 [rt.cpan.org #115553]
127 - in cygwin, drop the utime.t expected timestamp resolution
128 - documented for utime that the actual timestamp resolution depends
129 on the operating system and the filesystem, just like for stat
130 - looks like in gnukfreebsd 10.1 neither futimens or utimensat work
131 even though they exist, skipping utime tests there for now
132 - demodernize the Test::More use in utime.t (no subtest) to
133 better support older Perl installations
134 - for all the tests, drop the Test::More 0.82 requirement,
135 the only "modern" thing used is the note routine which
136 can be replaced with print("# ...\n") (or printf),
137 with this works back at least with Test::More 0.54 (perl 5.8.7)
140 - utime in HiRes.xs had dXSARGS which was unnecessary, leading
141 to two markstack pops, panic, and two-minute wait for the watchdog
142 (thanks to Dave Mitchell for the eagle-eye)
143 - even though 5.10.x should have the Perl_ck_warner, many CPAN
144 smokes proved otherwise, so requiring now at least 5.12 for
145 Perl_ck_warner use in IS_SAFE_PATHNAME()
148 - the IS_SAFE_PATHNAME() emulation was too modern for Perl 5.10
149 - make it so that only one value is set for -DTIME_HIRES_STAT
150 even on systems that support many options
151 - the 1.9734 (only released in blead, and as part of the development
152 release 5.25.2) fix for Darwin's clock_gettime was wrong under
153 threaded builds, leading into mutex panics in clock.t
154 - the d_hires_utime determination was made more precise
155 - the utime.t was failing in platforms where only one of
156 futimens/utimensat was present (neither or both was okay)
159 - Time::HiRes should override `utime` to allow setting hires
160 (futimens and utimensat support)
161 [rt.perl.org #114809]
162 - the utime patch used IS_SAFE_PATHNAME() which isn't available in
163 too old Perls, so emulate (in case the Devel::PPPort is too old)
166 - fix Darwins with clock_gettime: blead 2d41a263
167 [rt.perl.org #128427]
170 - C90 declaration-after-statement error with darwin threads: blead de1003b4
171 [rt.cpan.org #113856]
174 - MUTEX_LOCK, not PERL_MUTEX_LOCK: blead e5b02b5d
175 - also hrt_ualarm_itimer() is unused: 1cb6cce3
176 - the mutex needs init: 2d639e20
177 - version bump to 1.9732: 730d7fdc
180 - mark unused variable as such: blead a914236c
181 - OS X emulation mutex accidentally unused: da7a6455
182 - remove hrt_ualarm: 6da77c36
183 - pod error fixes: 919ca095
184 - nanosleep and clock_nanosleep cleanups: c8ea02b8..e3ff671b
185 - static funcs in HiRes.xs: 4e922c26
186 - Remove unused variable: 52ffb1b5
189 - TIME_HIRES_DONT_RUN_PROBES=1 to build the probes but not run them
190 [rt.cpan.org #111391]
194 [rt.cpan.org #111170]
195 - clock_getres(), clock_gettime(), clock_nanosleep() emulation for OS X
199 - No changes since 1.9727_03
201 1.9727_03 [2015-11-06]
202 - use PERL_UNUSED_ARG() in the not-here-croak branches
203 - Makefile.PL/metadata improvements
205 1.9727_02 [2015-08-29]
206 - correct documentation to note that Time::NVtime is NV, not double
208 - avoid a warning about clock_t on OS X / clang 3.6 - 3.6
209 [rt.cpan.org #102718]
211 1.9727_01 [2015-08-17]
212 - Remove usage of 5.005 thread APIs in favor of 5.6 and up thread APIs
213 [rt.cpan.org #102799].
214 - Silence a C++ compiler warning [rt.cpan.org #101497].
215 - respect $Config{run} for cross-compiling [rt.cpan.org #92607]
218 - Correct s/us splitting of usleep(1000000) [rt.cpan.org #78266].
219 - Avoid integer overflow in itimer-based alarm() with large
220 argument [rt.cpan.org #87160].
221 - Define PERL_NO_INLINE_FUNCTIONS during probe compilations, to
222 avoid false negatives caused by not linking with the perl core.
223 - Be more careful about context in stat().
224 - Install into "site" directories by default on perl 5.11+
225 [rt.cpan.org #79797].
226 - Fix a couple of doc typos [rt.cpan.org #85365].
227 - Fix function name in a doc example [rt.cpan.org #86318].
228 - Provide lstat() that yields high-res timestamps, alongside
229 the existing high-res stat() [rt.cpan.org #78732].
232 - Correct stack discipline in stat(), which was screwing up list
233 operations in expressions containing calls to it [rt.cpan.org
235 - Add missing OUTPUT sections to the XS code [rt.cpan.org #70930].
236 - Skip itimer tests on GNU/Hurd, which has the API but lacks
237 the implementation [rt.cpan.org #72754].
238 - Fix a doubled word in the documentation [rt.cpan.org #72763].
241 - Correct XS parameter list, and therefore prototype, for
242 unimplemented-on-this-platform version of clock_nanosleep()
243 [rt.cpan.org #68700].
244 - Declare package variables with "our" rather than "use vars".
245 - Corresponding to "our" usage, check for minimum Perl version
247 - Declare module dependencies.
250 - Remove $ENV{PERL_CORE} logic from test suite, which is no
251 longer desired in the core.
252 - Convert test suite to use Test::More.
253 - Factor out watchdog code from test suite.
254 - In test suite, be consistent about using fully-qualified form
256 - Divide test suite into feature-specific scripts.
257 - Make ualarm timing test less vulnerable to delay-induced false
258 failure, from Dave Mitchell.
261 - Update for changes in build process in the core, patches
262 from BinGOs [rt.cpan.org #58858] and Craig Berry [rt.cpan.org
264 - Fix broken linkage on Windows with gcc 3.4 seen with ActivePerl,
265 report from Christian Walde [rt.cpan.org #61648], fix derived
267 - Jump through hoops to avoid compiler warnings.
270 - Address [rt.cpan.org #54196] alarm and ularm return values are bogus,
271 additional fix from Gisle Aas
272 - Address [rt.cpan.org #55665] "Bad plan" on Windows,
273 report and fix from Jan Dubois
276 - Address [rt.cpan.org #54196] alarm and ularm return values are bogus,
277 report and fix from Nicholas Clark
280 - As with QNX, Haiku has the API of interval timers but not
281 the implementation (bleadperl change #34630), hence skip
282 the tests, via David Mitchell.
285 - .xs code cleanup from Albert Dvornik
286 - in the #39 and #40 do not do us I did, mixing alarm() and
287 sleep(). Now instead spin until enough time has passed.
290 - Skip the tests added in 1.9716 (#39, #40) if there's no subsecond
291 alarm capability, like with the older subsecond alarm tests
294 - Change documentation to agree with reality: there are
295 no interval timers in Win32.
296 - Address [rt.cpan.org #35899] (problem in subsecond sleeps),
297 add two tests to guard against this problem
298 - Address [rt.cpan.org #36600] 'Division by zero' failure in test suite
299 - Address [rt.cpan.org #37340] [PATCH] Address timer process in test
300 - Address [rt.cpan.org#40311 ] bad implementation of hrt_usleep
301 with TIME_HIRES_NANOSLEEP
304 - Silly me: Makefile.PL does need to accept arguments other than mine.
305 Some testing frameworks obviously do this.
306 - Add retrying for tests 34..37, which are the most commonly
307 failing tests. If this helps, consider extending the retry
308 framework to all the tests. [Inspired by Slaven Rezic,
309 [rt.cpan.org #34711] Occasional failures of test 35 or 36 (FreeBSD)]
312 - Under Perl 5.6.* NVgf needs to be "g", reported by Zefram,
313 it seems that ppport.h 3.13 gets this wrong.
314 - remove the check in Makefile.PL for 5.7.2, shouldn't be
315 (a) necessary (b) relevant
316 - add logic to Makefile.PL to skip configure/write Makefile
317 step if the "xdefine" file already exists, indicating that
318 the configure step has already been done, one can still
319 force (re)configure by "perl Makefile.PL configure",
320 or of course by "make clean && perl Makefile.PL".
323 - for alarm() and ualarm() [Perl] prefer setitimer() [C]
324 instead of ualarm() [C] since ualarm() [C] cannot portably
325 (and standards-compliantly) be used for more than 999_999
326 microseconds (rt.cpan.org #34655)
327 - it seems that HP-UX has started (at least in 11.31 ia64)
328 #defining the CLOCK_REALTIME et alia (instead of having
330 - document all the diagnostics
333 - move the sub tick in the test file back to where it used to be
334 - in the "consider upgrading" message recommend at least Perl 5.8.8
335 and make the message to appear only for 5.8.0 since 5.8.1 and
336 later have the problem fixed
337 - VOS tweak for Makefile (core perl change #33259)
338 - since the test #17 seems to fail often, relax its limits a bit
341 - lost VMS test skippage from Craig Berry
342 - reformat the test code a little
345 - I got the sense of the QNX test the wrong way in an attempt
346 to generalize it for future
349 - casting fixes from Robin Barker for g++ and 64bitint
350 - in QNX skip the itimer tests because though the API
351 is there, the implementation isn't, from Matt Kraai
352 - raise the dead man timer to 180 seconds for really
354 - elaborate the UTF-8 locale warning from Makefile.PL
357 - [rt.cpan.org #29747]: Build failure with perl 5.005_05
358 Fixed by regenerating the ppport.h using Devel::PPPort 3.13.
361 - t/HiRes.t failed in Perl 5.6.2,
362 "action is not of type POSIX::SigAction at t/HiRes.t line 318",
363 reported and fixed by Anton Berezin, the reason was faulty
364 use of sigaction() when restoring the old action.
367 - with bleadperl in VMS the HiRes.t overrun the maximum number
368 of deferred signals because the libc SIGALRM was not strong
369 enough to interrupt select(), and select() got restarted every
370 time, solution is to use POSIX::SigAction if available.
371 A fix from Craig Berry (not 100% there, but helps).
372 - allow for more measuring noise for ualarm() tests 35..37
375 - nanosleep() and clock_nanosleep() detection and use were
376 quite broken; in Linux -lrt needed; fixes from Zefram
377 - [internal] slightly cleaner building of $DEFINE in Makefile.PL,
378 should avoid double/conflicting -D flags
381 - allow 10% of slop in test #14 (testing difference between
382 CORE::time() and Time::HiRes::time()), there seem to be often
383 transient failures from Perl smoke builds on this test
387 - use int main(int argc, char **argv) consistently in Makefile.PL,
389 [rt.cpan.org #23868] nanosleep not detected under Mac OS 10.3.9 starting with Time::HiRes 1.96
390 - if someone still has the locale-broken Perl 5.8.0,
391 suggest that they upgrade their Perl
394 - restore the -DATLEASTFIVEOHOHFIVE, Win32 needed it still
397 - upgrade to ppport.h 3.10_02
398 - remove the -DATLEASTFIVEOHOHFIVE
399 - use the ppport.h PL_ppaddr, PL_statcache, PL_laststatval
400 - use the ppport.h aTHXR for calling Perl stat()
401 - switch into four-digit version since 2.0 is coming up
402 awfully fast but not feeling like a major rewrite
405 - 1.95 broke building in Win32 (since pp_stat is not exported),
406 figured out how to call an op directly in 5.005 (use Perl_ppaddr
407 instead of PL_ppaddr)
408 - backport to Perl 5.004_05 (requires using statcache
409 and laststatval instead of PL_statcache and PL_laststatval)
410 (also checked to work in 5.005_04, 5.6.1, and 5.8.8 with threads)
413 - 1.95 broke builds for threaded Perls, rt.cpan.org tickets:
414 [rt.cpan.org #23694] Time::HiRes fails tests on Solaris and Perl 5.6.1
415 [rt.cpan.org #23712] Time-HiRes 1.95 Fails make on AIX 5.2 with Perl 5.8.8
416 [rt.cpan.org #23730] Time::HiRes 1.95 fails make on MacOS X 10.3.9/perl 5.8.8
417 - use main() prototype consistently in Makefile.PL
420 - integrate core change #29180: Silence VC++ compiler warnings
422 - do not use PL_ppaddr in stat() because that is not available
424 - regenerate fallback/*.inc for older Perls without
425 ExtUtils::Constant because of d_hires_stat, resolves
426 [rt.cpan.org #23694] Time::HiRes fails tests on Solaris and Perl 5.6.1
427 - Make Makefile.PL more defensive against false PERL_CORE
430 - file timestamps oddities seen: the atime and mtime
431 can be out of sync (modify first and read second can leave
432 atime < mtime) and mtime can be subsecond while atime is not.
433 So make the test more forgiving.
436 - the ualarm() tests (34-37) assumed that ualarm(N)
437 could never alarm in less than N seconds, widened
438 the acceptable relative range to 0.9..1.5. Addresses
439 [rt.cpan.org #22090] and [rt.cpan.org #22091].
441 - skip the stat() tests in cygwin and win32, because
442 if run on FAT the timestamp granularity is only 2 seconds.
443 Any good way to detect (cygwin or win32) whether we are
444 being run on NTFS or anywhere with better timestamps?
445 Addresses [rt.cpan.org #22089] and [rt.cpan.org #22098].
448 - scan for subsecond resolution timestamps in struct stat,
449 some known possibilities:
451 (1) struct timespec st_atimespec;
452 st_atimespec.tv_nsec;
457 (4) timestruc_t st_atim;
462 If something like this is found, one can do
465 my @stat = Time::HiRes::stat();
467 or even override the standard stat():
469 use Time::HiRes qw(stat);
471 to get the stat() timestamps
473 my ($atime, $mtime, $ctime) = @stat[8, 9, 10];
475 with subsecond resolution (assuming both the operating
476 system and the filesystem support that kind of thing).
478 Contributions for more systems (especially non-UNIX,
479 e.g. but not limited to: Win32, VMS, OS/2) gladly accepted.
480 (also more UNIX variants welcome: HP-UX? IRIX?)
482 Thanks to H.Merijn Brand, John Peacock, and Craig
483 Berry for brave beta testing.
486 - ualarm() in SuSE 10.1 was overflowing after ~4.2 seconds,
487 possibly due to a glibc bug/feature (suspected overflow at
488 2**32 microseconds?), workaround by using the setitimer()
489 implementation of ualarm() if either useconds or
490 interval > 999_999 (this case seems to vary between systems:
491 are useconds more than 999_999 for ualarm() defined or not)
492 Added more ualarm() tests to catch various overflow points,
493 hopefully no problems in various platforms.
494 (The problem report by Mark Seger and Jon Paul Sullivan of HP.)
497 - tweak still needed for Const64(), from Jerry Hedden
498 - get a freshly generated ppport.h
499 - update Copyright years
502 - Const64() already appends an 'LL' (or i64), so provide LL and i64
503 forms for the IV_1E[679] (effects Win32 and Cygwin), reported by
505 - the Changes entry for 1.88 talked about [IN]V_1[679],
509 - clean up the g++ warnings in HiRes.xs, all of them
510 about mixing integer and floating point, introduce
511 constants IV_1E[679] and NV_1E[679]
514 - [rt.cpan.org #17442] 'make test' frequently fails under
515 Cygwin Perl v5.8.8, reported and patched by J. R. Hedden
516 (two race condition bugs in the END block in the case the
517 main process dies before the timer process, unearthed
518 by a bug in Cygwin ualarm)
521 - HiRes.t:s/ok 32/ok 33/, from Dominic Dunlop
522 - tighten up the clock() test marginally by requiring non-negative
523 - clock_nanosleep() and clock() doc tweaks
526 - the interface to clock_nanosleep() is more natural
527 when it is like (hires) time() (instead of like nanosleep),
528 and the .xs implementation of clock_nanosleep() in 1.84
530 - the semantics of clock() are not quite so silly as I thought,
531 but still somewhat odd, documented as such
532 - additional enhancements to the clock() documentation
533 - add test for clock_nanosleep() (I cannot test this
534 since none of my systems have the function)
535 - add test for clock()
538 - add clock() which returns the processor time in
539 (floating point) seconds since an arbitrary era
540 - add clock_nanosleep() which suspends the current
541 thread until either absolute time or for relative time
542 - [rt.cpan.org #16486] printf missing value in HiRes.t
543 - add constants CLOCKS_PER_SEC, CLOCK_SOFTTIME, TIMER_ABSTIME
547 - has_symbol() was wrong since e.g. ITIMER_VIRTUAL is exported
548 via @EXPORT_OK even when it is not available. This is heinous.
549 @EXPORT_OK should be determined at Makefile.PL time.
550 - be more lenient is testing clock_gettime(): allow more slop,
551 and retry up to three times, sleeping a random nap between
553 - human months are one-based (noticed by Anton Berezin)
556 - CLOCK_REALTIME is an enum value (of the clockid_t enum)
557 in HP-UX (and might be so elsewhere, too), debugged by
559 - include const-c.inc as late as possible (from Randy Kobes,
560 [rt.cpan.org #15552] to avoid undefined usleep() on Win32
563 - try to be more robust and consistent in the detection of
564 CLOCK_REALTIME and ITIMER_VIRTUAL in HiRes.t: the proper
569 eval 'import Time::HiRes qw($symbol)';
570 return 0 unless $@ eq '';
571 return exists ${"Time::HiRes::$symbol"};
578 in the test. All these moves are needed because
580 1) one cannot directly do eval 'Time::HiRes::FOO_BAR'
581 because FOO_BAR might have a true value of zero
582 (or in the general case an empty string or even undef)
584 2) In case FOO_BAR is not available in this platform,
585 &FOO_BAR avoids the bareword warning
587 - wait more (1.5 seconds instead of 0.1) for the CLOCK_REALTIME test
588 but expect the 'customary' slop of 0.20 instead of 0.25
589 - fixed inside a comment HAS_POLL -> TIME_HIRES_NANOSLEEP
590 - at the end of HiRest.t tell how close we were to termination
593 - Gisle noticed a mistake (using HAS_NANOSLEEP) in 1.79
596 - try nanosleep for emulating usleep -- may help in some weird
597 embedded realtime places which have nanosleep but neither usleep
598 nor select nor poll (doesn't have to be weird embedded realtime
599 place, though -- in many places usleep is nanosleep anyway)
600 - try poll for emulating usleep -- this may help some obscure/old
601 SVR4 places that have neither usleep nor select
602 - a redundant test guard in HiRes.t
605 - ITIMER_VIRTUAL detection in HiRes.t had problems (that we cannot
606 in the general case fail already at 'use' phase is suboptimal)
607 - fixes to the documentation of clock_gettime() and clock_getres()
610 - add support for the POSIX clock_gettime() and clock_getres(),
611 if available, either as library calls or as syscalls
612 - be more defensive about missing functionality: break out
613 early (during 'use') if no e.g. clock_getres() is available,
614 and protect our back by trapping those cases also in HiRes.xs
615 - the test added in 1.76 could cause an endless loop e.g. in Solaris,
616 due to mixing of sleep() and alarm() (bad programmer, no cookie!)
619 - testing for nanosleep had wrong logic which caused nanosleep
620 to become undefined for e.g. Mac OS X
621 - added a test for a core dump that was introduced by Perl 5.8.0
622 safe signals and was fixed for the time of 5.8.1 (one report of
623 the core dump was [perl #20920]), the test skipped pre-5.8.1.
624 - *cough* s/unanosleep/nanosleep/g; *cough*
627 - installation patch from Gisle Aas: in Perls 5.8.x and later
628 use MakeMaker INSTALLDIRS value of 'perl' instead of 'site'.
631 - [cpan #14608] Solaris 8 perl 5.005_03 File::Spec module does not have method rel2abs
632 (the workaround is not to use rel2abs, should not be necessary)
633 - [cpan #14642] U2time wrongly exported on the C API
634 (patch supplied by the reporter, SALVA@cpan.org)
635 - add release dates to Changes
638 - Time::HiRes::nanosleep support for Solaris [PATCH]
639 (POSIX::uname() not available if building with core perl,
640 from Gisle Aas, via perl5-porters, perl change #25295)
643 - going back to the 1.68 loader setup (using DynaLoader)
644 since too many weird things starting breaking
645 - fix a typo in José Auguste-Etienne's name
648 - a thinko in the nanosleep() detection
649 - move more changes stuff from the README to Changes
650 - add -w to the Makefile.PL
653 - oops in 1.69 about @ISA (not affecting anything but silly)
654 - add copyright 2005 to HiRes.pm
655 - add copyright and license to HiRes.xs
656 - add copyrights 2003, 2004, 2005 to README
659 - actually run a test for nanosleep
660 (if there is no $Config{d_nanosleep}) since e.g. in AIX 4.2
661 it seems that one can link in nanosleep() but then calling
662 it fails instantly and sets errno to ENOSYS (Not implemented).
663 This may be fixable in the AIX case by figuring out the right
664 (realtime POSIX?) libs and whatnot, but in the general case
665 running a real test case is better. (Of course, this change
666 will no doubt run into portability problems because of the
667 execution step...) Note that because of hysterical raisins
668 most Perls do NOT have $Config{d_nanosleep} (scanning for
669 it by Configure would in many platforms require linking in
670 things like -lrt, which would in many platforms be a bad idea
672 (from José Auguste-Etienne)
673 - support XSLoader also since it's much faster
674 (from Alexey Tourbin)
675 - add SEE ALSO (BSD::Resource and Time::TAI64)
678 - somehow 1.67 had a lot of doubled lines (a major cut-and-paste
679 error suspected), but miraculously it still worked since the
680 doubling took place below the __END__ token
681 - undef Pause() before defining it to avoid redefinition warnings
682 during compilation in case perl.h had already defined Pause()
683 (part of perl change #24271)
687 - (internal) don't ignore the return value of gettimeofday()
688 - (external) return undef or an empty if the C gettimeofday() fails
689 (affects Time::HiRes gettimeofday() and the hires time())
693 - fix the 'hierachy' typo in Makefile.PL [rt.cpan.org #8492]
694 - should now build in Solaris [rt.cpan.org #7165] (since 1.64)
695 - should now build in Cygwin [rt.cpan.org #7535] (since 1.64)
696 - close also [rt.cpan.org #5933] "Time::HiRes::time does not
697 pick up time adjustments like ntp" since ever reproducing it
698 (and therefore verifying a possible fix) in the same environment
699 has become rather unlikely
702 - one should not mix u?alarm and sleep (the tests modified
703 by 1.65, #12 and #13, hung in Solaris), now we just busy
704 loop executing an empty block
705 - in the documentation underline the unspecificity of mixing
707 - small spelling fixes
710 - regenerate ppport.h with Devel::PPPort 3.03,
711 now the MY_CXT_CLONE is defined in ppport.h,
712 we no more need to do that.
714 - the test #12 would often hang in sigsuspend() (at least that's
715 where Mac OS X' ktrace shows it hanging). With the sleep()s
716 changed to sleep(1)s, the tests still pass but no hang after
717 a few hundred repeats.
720 - Win32 and any ithread build: ppport.h didn't define
721 MY_CXT_CLONE, which seems to be a Time-HiRes-ism.
724 - Skip testing if under PERL_CORE and Time::HiRes has not
725 been Configured (from Marcus Holland-Moritz, core change
727 - Use ppport.h generated by Devel::PPPort 3.01,
728 allowing cutting away our own portability code.
729 - Don't use $ENV{PERL_CORE} for < 5.6.0.
730 - Don't use "for my $i" for <= 5.003.
731 - Don't use Pause() for <= 5.003.
732 - Can't use newSVpvf for <= 5.003.
733 (most of the changes from Marcus)
736 - Win32: reset reading from the performance counters every
737 five minutes to better track wall clock time (thanks to
738 PC timers being often quite bad), should help long-running
742 - Win32: Patch from Steve Hay
743 [PATCH] Re: [perl #30755] [Win32] Different results from Time::HiRes::gettimeofdayunder the debugger
744 to [perl #30755] reported by Nigel Sandever
746 - Cygwin: Use the Win32 recalibration code also in Cygwin if the
747 <w32api/windows.h> APIs are available. Cygwin testing by
748 Yitzchak Scott-Thoennes.
750 - Solaris: use -lposix4 to get nanosleep for Solaris 2.6,
751 after that keep using -lrt, patch from Alan Burlison,
752 bug reported in [cpan #7165]
755 - Change the Win32 recalibration limit to 0.5 seconds and tweak
756 the documentation to blather less about the gory details of the
757 Win32 implementation and more about the complications in general
758 of meddling with the system clock.
761 - Document the 1.57 change better.
764 - Win32/Cygwin/MinGW: if the performance counter drifts by more
765 than two seconds from the system clock (due to ntp adjustments,
766 for example), recalibrate our internal counter: from Jan Dubois,
767 based on [cpan #5933] by Jerry D. Hedden.
770 - Give a clearer message if the tests timeout (perl change #22253)
771 - Don't use /tmp or its moral equivalents (perl bug #15036,
775 - Windows: mingw32 patch from Mike Pomraning (use Perl's Const64()
776 instead of VC-specific i64 suffix)
779 - Solaris: like Tru64 (dec_osf) also Solaris need -lrt for nanosleep
782 - Windows: higher resolution time() by using the Windows
783 performance counter API, from Jan Dubois and Anton Shcherbinin.
784 The exact new higher resolution depends on the hardware,
785 but it should be quite a bit better than using the basic
789 - In AIX (v?) with perl 5.6.1 the HiRes.t can hang after
790 the subtest 18. No known analysis nor fix, but added
791 an alarm (that requires fork() and alarm()) to the test.
794 - doc tweaks from mjd (perl change #20456)
795 - NCR MP-RAS hints file added (svr4.pl) (perl change #21249)
798 - add a message (for non-core builds) to Makefile.PL about
799 the LC_ALL=C workaround
800 - &Time::HiRes::d_nanosleep was broken (perl change #20131)
801 - the nanosleep() probe was broken (perl change #20061)
802 - use existence instead of definedness for feature probes
804 - MPE/iX tweak (perl change #20042)
805 - do not use HAS_NANOSLEEP (perl change #19898)
808 - UVuf for non-IVSIZE platforms (from Keiichiro Nagano)
809 - OS/2 can always mix subsecond sleeps with signals
810 (part of perl change #19789)
813 - workaround for buggy gcc 2.95.3 in openbsd/sparc64
817 - do not use -lrt in Linux (from March Lehmann, perl change #19449)
818 - unnecessary (nanosleep is in libc anyway)
819 - harmful (-lrt slows down execution)
820 - incompatible (with many distributions' pthreads)
823 - do not create files in blib directories under core
824 (perl change #19160, from rgs)
825 - detypo s/VTLARM/VTARLM/ (perl change #19328, from mjd)
828 - guarantee that $xdefine in HiRes.t is always defined
829 (perl change #19109, from IlyaZ)
830 - a cleaner way to detect PERL_CORE (perl change #19111,
834 - add hints/irix.pl to turn off overly POSIX flags that
835 cause hide struct timespec to be hidden (and compilation
836 to fail) (bleadperl change #19085)
837 - documentation tweaks
840 - add c:/temp to the list of temp directories to probe
841 so that cygwin (and win*?) builds are happy. This was
842 needed at least in my cygwin 1.3.20/w2k setup.
845 - modernize the constants code (from Nicholas Clark)
848 - At some point the ability to figure our the correct incdir
849 for EXTERN.h (either a core perl build, or an installed perl)
850 had broken (which lead into all test compiles failing with
851 a core perl build, but thanks to the robustness of Makefile.PL
852 nothing of this was visible). The brokenness seemed to be
853 caused by $ENV{PERL_CORE} not being on for core builds?
854 Now stole a trick from the Encode that sets $ENV{PERL_CORE}
855 right, and both styles of build should work again.
858 - Nicholas Clark noticed that the my_catdir() emulation function
859 was broken (which means that we didn't really work for Perls
861 - inspired by fixing the above made the whole Makefile.PL -w
863 - tightened up the Makefile.PL output, less whitespace
866 - fix from Craig Berry for better building in VMS with PERL_CORE
869 - no functional changes
870 - move lib/Time/HiRes.pm as Hires.pm
871 - libraries scanning was slightly broken (always scanned
872 for a library even when $Config{libs} already had it)
875 - Ray Zimmerman ran into a race condition in Mac OS X.
876 A 0.01-second alarm fired before the test expected.
877 The test first slept indefinitely (blocking for signals)
878 and only after that tested for the signal having been sent.
879 Since the signal had already been sent, the test #12 never
880 completed. The solution: test first, then block.
881 - default to being silent on all probing attempts, set the
882 environment variable VERBOSE to a true value to see the
883 details (the probing command and the possible errors)
886 - do not clear MAN3PODS in Makefile.PL (Radoslaw Zielinski)
887 - INSTALLDIRS => 'perl' missing which means that Time::HiRes
888 cannot be upgraded from CPAN to override the 5.8.0 version
890 - Time::HiRes 1.35 could not be dropped as-is to bleadperl
891 because the include directories did not adjust themselves
892 if $ENV{PERL_CORE} (Hugo van der Sanden)
893 - add documentation about the restart of select() under alarm()
896 - small documentation tweaks
900 - better VMS operation (Craig Berry)
903 - our time machine is accelerating: now works with Perl 5.004_01
904 (tried with 5.003_07 and 5.002 but I get segmentation faults
905 from running the Makefile.PL with those in Tru64 4.0D)
908 - backward compatibility (pre-5.6.0) tweaks:
909 - no XSLoader in 5.00503, use DynaLoader instead
910 - no SvPV_nolen, either
911 - no PerlProc_pause(), either
912 - now tested with 5.00404 and 5.00503
913 - Makefile.PL requires 5.00404 (no more 5.002)
914 - use nanosleep instead of usleep, if it is available (Wilson Snyder)
915 (this means that one can mix subsecond sleeps with alarms)
916 - because of nanosleep we probe for -lrt and -lposix4
917 - the existence of getitimer/nanosleep/setitimer/ualarm/usleep
918 is available by exportable constants Time::HiRes::d_func
919 (since older Perl do not have them in %Config, and even
920 5.8.0 does not probe for nanosleep)
923 - backward compatibility (pre-5.6.1) tweaks:
924 - define NV if no NVTYPE
925 - define IVdf if needed (note: the Devel::PPPort
926 in 5.8.0 does not try hard hard enough since
927 the IVSIZE might not be defined)
928 - define NVgf if needed
929 - grab the typemap from 5.8.0 for the NV stuff
931 1.31 and 1.32 add more backward compatibility (now all the way
932 back to Perl 5.00404), and using nanosleep() (if available) for
937 - release 1.29_02 as 1.30
939 1.30 adds all the changes made during the Perl 5.6->5.7->5.8
940 development cycle. Most notably portability across platforms has been
941 enhanced, and the interval timers (setitimer, getitimer) have been
942 added. Note that the version of Time::HiRes that is included in Perl
943 5.8.0 calls itself 1.20_00, but it is equivalent to this Time::HiRes
944 version. Note also that in 1.30 Wegscheid turns over the maintenance
945 to Jarkko Hietaniemi.
949 - fix a silly unclosed comment typo in HiRes.xs
950 - document and export REALTIME_REALPROF (Solaris)
954 - only getitimer(ITIMER_REAL) available in Cygwin and Win32
955 (need to patch this also in Perl 5.[89])
956 - remove CVS revision log from HiRes.xs
960 The following numbered patches refer to the Perl 5.7 changes,
961 you can browse them at http://public.activestate.com/cgi-bin/perlbrowse
963 - 17558: Add #!./perl to the .t
964 - 17201: linux + usemorebits fix, from Rafael Garcia-Suarez
965 - 16198: political correctness, from Simon Cozens
966 - 15857: doc tweaks, from Jarkko Hietaniemi
967 - 15593: optimization in .xs, from Paul Green
968 - 14892: pod fixes, from Robin Barker
969 - 14100: VOS fixes, from Paul Green
970 - 13422: XS segfault, from Marc Lehmann
971 - 13378: whether select() gets restarted on signals, depends
972 - 13354: timing constraints, again, from Andy Dougherty
973 - 13278: can't do subsecond alarms with ualarm;
974 break out early if alarms do not seem to be working
975 - 13266: test relaxation (cygwin gets lower hires
976 times than lores ones)
977 - 12846: protect against high load, from Jarkko Hietaniemi
978 - 12837: HiRes.t VMS tweak, from Craig A. Berry
979 - 12797: HiRes.t VMS tweak, from Charles Lane
980 - 12769: HiRes.t VMS tweak, from Craig A. Berry
981 - 12744: gcc vs MS 64-bit constant syntax, from Nick Ing-Simmons
982 - 12722: VMS ualarm for VMS without ualarm, from Charles Lane
983 - 12692: alarm() ain't gonna work if ualarm() ain't,
984 from Gurusamy Sarathy
985 - 12680: minor VMS tweak, from Charles Lane
986 - 12617: don't try to print ints as IVs, from Jarkko Hietaniemi
987 - 12609: croak on negative time, from Jarkko Hietaniemi
988 - 12595: Cygwin rounds up for time(), from Jarkko Hietaniemi
989 - 12594: MacOS Classic timeofday, from Chris Nandor
990 - 12473: allow for more than one second for sleep() and usleep()
991 - 12458: test tuning, relax timing constraints,
992 from Jarkko Hietaniemi
993 - 12449: make sleep() and usleep() to return the number
994 of seconds and microseconds actually slept (analogously
995 with the builtin sleep()), also make usleep() croak if
996 asked for more than 1_000_000 useconds, from Jarkko Hietaniemi
997 - 12366: Time::HiRes for VMS pre-7.0, from Charles Lane
998 - 12199: do not use ftime on Win32, from Gurusamy Sarathy
999 - 12196: use ftime() on Win32, from Artur Bergman
1000 - 12184: fix Time::HiRes gettimeofday() on Win32, from Gurusamy Sarathy
1001 - 12105: use GetSystemTime() on Win32, from Artur Bergman
1002 - 12060: explain the 1e9 seconds problem, from Jarkko Hietaniemi
1003 - 11901: UNICOS sloppy division, from Jarkko Hietaniemi
1004 - 11797: problem in HiRes.t, from John P. Linderman
1005 - 11414: prototype from Time::HiRes::sleep(), from Abhijit Menon-Sen
1006 - 11409: Time::HiRes qw(sleep) failed, from Abhijit Menon-Sen
1007 - 11270: dynix/ptx 4.5.2 hints fix, from Peter Prymmer
1008 - 11032: VAX VMS s/div/lib\$ediv/ fix, from Peter Prymmer
1009 - 11011: VAX VMS s/qdiv/div/ fix, from Peter Prymmer
1010 - 10953: SCO OpenServer 5.0.5 requires an explicit -lc for usleep(),
1012 - 10942: MPE/IX test tweaks, from Mark Bixby
1013 - 10784: unnecessary pod2man calls, from Andy Dougherty
1014 - 10354: ext/ + -Wall, from Doug MacEachern
1015 - 10320: fix the BOOT section to call myU2time correctly
1016 - 10317: correct casting for AIX< from H. Merijn Brand
1017 - 10119: document that the core time() may be rounding, not truncating
1018 - 10118: test fix, from John Peacock
1019 - 9988: long =item, from Robin Barker
1020 - 9714: correct test output
1021 - 9708: test also the scalar aspect of getitimer()
1022 - 9705: Add interval timers (setitimer, getitimer)
1023 - 9692: do not require at least 5.005 using XS
1025 The following changes were made on top of the changes
1026 made for Time::HiRes during the Perl 5.7 development
1027 cycle that culminated in the release of Perl 5.8.0.
1029 - add "require 5.005" to the Makefile.PL
1030 - remove the REVISION section (CVS log) from HiRes.pm
1031 - add jhi's copyright alongside Douglas'
1032 - move HiRes.pm to lib/Time/
1033 - move HiRes.t to t/
1034 - modify HiRes.t to use $ENV{PERL_CORE}
1035 - modify the original Time::HiRes version 1.20 Makefile.PL
1036 to work both with Perl 5.8.0 and the new code with pre-5.8.0
1037 Perls (tried with 5.6.1)
1038 - tiny tweaks and updates in README and TODO
1039 - bump the VERSION to 1.29
1041 1.20 Wed Feb 24 21:30 1999
1042 - make our usleep and ualarm substitutes into hrt_usleep
1043 and hrt_ualarm. This helps static links of Perl with other
1044 packages that also have usleep, etc. From
1045 Ilya Zakharevich <ilya@math.ohio-state.edu>
1046 - add C API stuff. From Joshua Pritikin
1047 <joshua.pritikin@db.com>
1048 - VMS Makefile.PL fun. From pvhp@forte.com (Peter Prymmer)
1049 - hopefully correct "-lc" fix for SCO.
1052 1.20 adds a platform neutral set of C accessible routines if you are
1053 running 5.005+. All other changes are packaging changes and build
1054 fixes(?) for statically linked Perl, SCO, and VMS.
1056 1.19 Tue Sep 29 22:30 1998
1057 - put VMS gettimeofday() in. Patch is from Sebastian Bazley
1058 <seb@stian.demon.co.uk>
1059 - change GIMME_V to GIMME to help people with older versions of
1061 - fix Win32 version of gettimeofday(). It didn't affect anything,
1062 but it confuses people reading the code when the return value
1063 is backwards (0 is success).
1064 - fix Makefile.PL (more) so that detection of gettimeofday is
1067 1.19 has better VMS support.
1069 1.18 Mon Jul 6 22:40 1998
1070 - add usleep() for Win32.
1071 - fix Makefile.PL to fix reported HP/UX feature where unresolved
1072 externals still cause an executable to be generated (though no
1073 x bit set). Thanks to David Kozinn for report and explanation.
1074 Problems with the fix are mine :)
1076 1.18 has limited Win32 support (no ualarm). Added usleep for Win32.
1077 Probably buggy. I'm sure I'll hear.
1079 1.17 Wed Jul 1 20:10 1998
1080 - fix setitimer calls so microseconds is not more than 1000000.
1081 Hp/UX 9 doesn't like that. Provided by Roland B Robert, PhD.
1082 - make Win32. We only get gettimeofday (the select hack doesn't
1083 seem to work on my Win95 system).
1084 - fix test 4 on 01test.t. add test to see if time() and
1085 Time::HiRes::time() are close.
1087 1.16 Wed Nov 12 21:05 1997
1088 - add missing EXTEND in new gettimeofday scalar code.
1090 1.16+ should be closer to building out of the box on Linux. Thanks
1091 to Gisle Aas for patches, and the ualarm equivalent using setitimer.
1093 If your underlying operating system doesn't implement ualarm(), then
1094 a fake using setitimer() will be made. If the OS is missing usleep(),
1095 a fake one using select() will be made. If a fake can't be made for
1096 either ualarm() or usleep(), then the corresponding Perl function will
1097 not be available. If the OS is missing gettimeofday(), you will get
1098 unresolved externals, either at link- or run-time.
1100 This is an improvement; the package used to not even build if
1101 you were missing any of these bits. Roderick Schertler
1103 <roderick@argon.org> did all the conditional compilation stuff,
1104 look at HiRes.pm and the test suites; it's good educational reading.
1106 1.15 Mon Nov 10 21:30 1997
1107 - HiRes.pm: update pod. Provided by Gisle Aas.
1108 - HiRes.xs: if gettimeofday() called in scalar context, do
1109 something more useful than before. Provided by Gisle Aas.
1110 - README: tell of xsubpp '-nolinenumber' woes. thanks to
1111 Edward Henigin <ed@texas.net> for pointing out the problem.
1113 1.14 Wed Nov 5 9:40 1997
1114 - Makefile.PL: look for setitimer
1115 - HiRes.xs: if missing ualarm, but we have setitimer, make up
1116 our own setitimer. These were provided by Gisle Aas.
1118 1.13 Tue Nov 4 23:30 1997
1119 - Makefile.PL: fix autodetect mechanism to do try linking in addition
1120 to just compiling; should fix Linux build problem. Fix was provided
1123 1.12 Sun Oct 12 12:00:00 1997
1124 - Makefile.PL: set XSOPT to '-nolinenumbers' to work around xsubpp bug;
1125 you may need to comment this back out if you have an older xsubpp.
1126 - HiRes.xs: set PROTOTYPES: DISABLE
1128 1.11 Fri Sep 05 16:00:00 1997
1130 Had some line commented out that shouldn't have been (testing
1133 Previous version was corrupted.
1135 1.10 Thu May 22 20:20:00 1997
1136 - HiRes.xs, HiRes.pm, t/*:
1137 - only compile what we have OS support for (or can
1139 - only test what we compiled
1140 - gross improvement to the test suite
1142 This work was all done by Roderick Schertler
1143 <roderick@argon.org>. If you run Linux or
1144 one of the other ualarm-less platforms, and you like this
1145 module, let Roderick know; without him, it still wouldn't
1146 be working on those boxes...
1147 - Makefile.PL: figure out what routines the OS has and
1148 only build what we need. These bits were written by Jarkko
1149 Hietaniemi <jhi@iki.fi>. Again, gratitude is due...
1151 1.02 Mon Dec 30 08:00:00 1996
1152 - HiRes.pm: update documentation to say what to do when missing
1153 ualarm() and friends.
1154 - README: update to warn that ualarm() and friends need to exist
1156 1.01 Fri Oct 17 08:00:00 1996
1157 - Makefile.PL: make XSPROTOARGS => '-noprototyopes'
1158 - HiRes.pm: put blank line between __END__ and =head1 so that
1161 1.00 Tue Sep 03 13:00:00 1996
1162 - original version; created by h2xs 1.16