Changes   [plain text]

Revision history for Perl extension Time::HiRes.

	- Change the Win32 recalibration limit to 0.5 seconds and tweak
	  the documentation to blather less about the gory details of the
	  Win32 implementation and more about the complications in general
	  of meddling with the system clock.

	- Document the 1.57 change better.

	- Win32/Cygwin/MinGW: if the performance counter drifts by more
	  than two seconds from the system clock (due to ntp adjustments,
	  for example), recalibrate our internal counter: from Jan Dubois,
	  based on [cpan #5933] by Jerry D. Hedden.

	- Give a clearer message if the tests timeout (perl change #22253)
	- Don't use /tmp or its moral equivalents (perl bug #15036,
	  perl change #22258)

	- Windows: ming32 patch from Mike Pomraning (use Perl's Const64()
	  instead of VC-specific i64 suffix)

	- Solaris: like Tru64 (dec_osf) also Solaris need -lrt for nanosleep

	- Windows: higher resolution time() by using the Windows
	  performance counter API, from Jan Dubois and Anton Shcherbinin.
	  The exact new higher resolution depends on the hardware,
	  but it should be quite a bit better than using the basic
	  Windows timers.

	- In AIX (v?) with perl 5.6.1 the HiRes.t can hang after
	  the subtest 18.  No known analysis nor fix, but added
	  an alarm (that requires fork() and alarm()) to the test.

	- doc tweaks from mjd (perl change #20456)
	- NCR MP-RAS hints file added ( (perl change #21249)

	- add a message (for non-core builds) to Makefile.PL about
	  the LC_ALL=C workaround
	- &Time::HiRes::d_nanosleep was broken (perl change #20131)
	- the nanosleep() probe was broken (perl change #20061)
	- use existence instead of definedness for feature probes
	  (perl change #20043)
	- MPE/iX tweak (perl change #20042)
	- do not use HAS_NANOSLEEP (perl change #19898)

	- UVuf for non-IVSIZE platforms (from Keiichiro Nagano)
	- OS/2 can always mix subsecond sleeps with signals
	  (part of perl change #19789)

	- workaround for buggy gcc 2.95.3 in openbsd/sparc64
	  (perl change #19592)

	- do not use -lrt in Linux (from March Lehmann, perl change #19449)
		- unnecessary (nanosleep is in libc anyway)
		- harmful (-lrt slows down execution)
		- incompatible (with many distributions' pthreads)

	- do not create files in blib directories under core
	  (perl change #19160, from rgs)
	- detypo s/VTLARM/VTARLM/ (perl change #19328, from mjd)

	- guarantee that $xdefine in HiRes.t is always defined
	  (perl change #19109, from IlyaZ)
	- a cleaner way to detect PERL_CORE (perl change #19111,
	  from IlyaZ)

	- add hints/ to turn off overly POSIX flags that
	  cause hide struct timespec to be hidden (and compilation
	  to fail) (bleadperl change #19085)
	- documentation tweaks

	- add c:/temp to the list of temp directories to probe
	  so that cygwin (and win*?) builds are happy.  This was
	  needed at least in my cygwin 1.3.20/w2k setup.

	- modernize the constants code (from Nicholas Clark)

	- At some point the ability to figure our the correct incdir
	  for EXTERN.h (either a core perl build, or an installed perl)
	  had broken (which lead into all test compiles failing with
	  a core perl build, but thanks to the robustness of Makefile.PL
	  nothing of this was visible).  The brokenness seemed to be
	  caused by $ENV{PERL_CORE} not being on for core builds?
	  Now stole a trick from the Encode that sets $ENV{PERL_CORE}
	  right, and both styles of build should work again.

	- Nicholas Clark noticed that the my_catdir() emulation function
	  was broken (which means that we didn't really work for Perls
	  5.002 and 5.003)
	- inspired by fixing the above made the whole Makefile.PL -w
	  and strict clean
	- tightened up the Makefile.PL output, less whitespace

	- fix from Craig Berry for better building in VMS with PERL_CORE

	- no functional changes
	- move lib/Time/ as
	- libraries scanning was slightly broken (always scanned
	  for a library even when $Config{libs} already had it)

	- Ray Zimmerman ran into a race condition in Mac OS X.
	  A 0.01-second alarm fired before the test expected.
	  The test first slept indefinitely (blocking for signals)
	  and only after that tested for the signal having been sent.
	  Since the signal had already been sent, the test #12 never
	  completed.  The solution: test first, then block.
	- default to being silent on all probing attempts, set the
	  environment variable VERBOSE to a true value to see the
	  details (the probing command and the possible errors)

	- do not clear MAN3PODS in Makefile.PL (Radoslaw Zielinski)
	- INSTALLDIRS => 'perl' missing which means that Time::HiRes
	  cannot be upgraded from CPAN to override the 5.8.0 version
	  (Guido A. Ostkamp)
	- Time::HiRes 1.35 could not be dropped as-is to bleadperl
	  because the include directories did not adjust themselves
	  if $ENV{PERL_CORE} (Hugo van der Sanden)
	- add documentation about the restart of select() under alarm()

	- small documentation tweaks

	- better VMS operation (Craig Berry)

	- our time machine is accelerating: now works with Perl 5.004_01
	  (tried with 5.003_07 and 5.002 but I get segmentation faults
	   from running the Makefile.PL with those in Tru64 4.0D)

	- backward compatibility (pre-5.6.0) tweaks:
	  - no XSLoader in 5.00503, use DynaLoader instead
	  - no SvPV_nolen, either
	  - no PerlProc_pause(), either
	  - now tested with 5.00404 and 5.00503
	  - Makefile.PL requires 5.00404 (no more 5.002)
	- use nanosleep instead of usleep, if it is available (Wilson Snyder)
	  (this means that one can mix subsecond sleeps with alarms)
	- because of nanosleep we probe for -lrt and -lposix4
	- the existence of getitimer/nanosleep/setitimer/ualarm/usleep
	  is available by exportable constants Time::HiRes::d_func
	  (since older Perl do not have them in %Config, and even
	   5.8.0 does not probe for nanosleep)

	- backward compatibility (pre-5.6.1) tweaks:
	  - define NV if no NVTYPE
	  - define IVdf if needed (note: the Devel::PPPort
	    in 5.8.0 does not try hard hard enough since
	    the IVSIZE might not be defined)
	  - define NVgf if needed
	  - grab the typemap from 5.8.0 for the NV stuff

	- release 1.29_02 as 1.30


	- fix a silly unclosed comment typo in HiRes.xs
	- document and export REALTIME_REALPROF (Solaris)


	- only getitimer(ITIMER_REAL) available in Cygwin and Win32
	  (need to patch this also in Perl 5.[89])
	- remove CVS revision log from HiRes.xs


	The following numbered patches refer to the Perl 5.7 changes,
	you can browse them at

	- 17558: Add #!./perl to the .t
	- 17201: linux + usemorebits fix, from Rafael Garcia-Suarez
	- 16198: political correctness, from Simon Cozens
	- 15857: doc tweaks, from Jarkko Hietaniemi
	- 15593: optimization in .xs, from Paul Green
	- 14892: pod fixes, from Robin Barker
	- 14100: VOS fixes, from Paul Green
	- 13422: XS segfault, from Marc Lehmann
	- 13378: whether select() gets restarted on signals, depends
	- 13354: timing constraints, again, from Andy Dougherty
	- 13278: can't do subecond alarms with ualarm;
		 break out early if alarms do not seem to be working
	- 13266: test relaxation (cygwin gets lower hires
		 times than lores ones)
	- 12846: protect against high load, from Jarkko Hietaniemi
	- 12837: HiRes.t VMS tweak, from Craig A. Berry
	- 12797: HiRes.t VMS tweak, from Charles Lane
	- 12769: HiRes.t VMS tweak, from Craig A. Berry
	- 12744: gcc vs MS 64-bit constant syntax, from Nick Ing-Simmons
	- 12722: VMS ualarm for VMS without ualarm, from Charles Lane
	- 12692: alarm() ain't gonna work if ualarm() ain't,
		 from Gurusamy Sarathy
	- 12680: minor VMS tweak, from Charles Lane
	- 12617: don't try to print ints as IVs, from Jarkko Hietaniemi
	- 12609: croak on negative time, from Jarkko Hietaniemi
	- 12595: Cygwin rounds up for time(), from Jarkko Hietaniemi
	- 12594: MacOS Classic timeofday, from Chris Nandor 
	- 12473: allow for more than one second for sleep() and usleep()
	- 12458: test tuning, relax timing constraints,
		 from Jarkko Hietaniemi
	- 12449: make sleep() and usleep() to return the number
		 of seconds and microseconds actually slept (analogously
		 with the builtin sleep()), also make usleep() croak if
		 asked for more than 1_000_000 useconds, from Jarkko Hietaniemi
	- 12366: Time::HiRes for VMS pre-7.0, from Charles Lane
	- 12199: do not use ftime on Win32, from Gurusamy Sarathy
	- 12196: use ftime() on Win32, from Artur Bergman
	- 12184: fix Time::HiRes gettimeofday() on Win32, from Gurusamy Sarathy
	- 12105: use GetSystemTime() on Win32, from Artur Bergman
	- 12060: explain the 1e9 seconds problem, from Jarkko Hietaniemi
	- 11901: UNICOS sloppy division, from Jarkko Hietaniemi
	- 11797: problem in HiRes.t, from John P. Linderman
	- 11414: prototype from Time::HiRes::sleep(), from Abhijit Menon-Sen
	- 11409: Time::HiRes qw(sleep) failed, from Abhijit Menon-Sen
	- 11270: dynix/ptx 4.5.2 hints fix, from Peter Prymmer 
	- 11032: VAX VMS s/div/lib\$ediv/ fix, from Peter Prymmer
	- 11011: VAX VMS s/qdiv/div/ fix, from Peter Prymmer
	- 10953: SCO OpenServer 5.0.5 requires an explicit -lc for usleep(),
		 from Jonathan Stowe
	- 10942: MPE/IX test tweaks, from Mark Bixby
	- 10784: unnecessary pod2man calls, from Andy Dougherty 
	- 10354: ext/ + -Wall, from Doug MacEachern
	- 10320: fix the BOOT section to call myU2time correctly
	- 10317: correct casting for AIX< from H. Merijn Brand
	- 10119: document that the core time() may be rounding, not truncating
	- 10118: test fix, from John Peacock
	-  9988: long =item, from Robin Barker
	-  9714: correct test output
	-  9708: test also the scalar aspect of getitimer()
	-  9705: Add interval timers (setitimer, getitimer)
	-  9692: do not require at least 5.005 using XS
	The following changes were made on top of the changes
	made for Time::HiRes during the Perl 5.7 development
	cycle that culminated in the release of Perl 5.8.0. 

	- add "require 5.005" to the Makefile.PL
	- remove the REVISION section (CVS log) from
	- add jhi's copyright alongside Douglas'
	- move to lib/Time/
	- move HiRes.t to t/
	- modify HiRes.t to use $ENV{PERL_CORE}
	- modify the original Time::HiRes version 1.20 Makefile.PL
	  to work both with Perl 5.8.0 and the new code with pre-5.8.0
	  Perls (tried with 5.6.1)
	- tiny tweaks and updates in README and TODO
	- bump the VERSION to 1.29

1.20  Wed Feb 24 21:30 1999
	- make our usleep and ualarm substitutes into hrt_usleep 
	  and hrt_ualarm. This helps static links of Perl with other
	  packages that also have usleep, etc. From
	  Ilya Zakharevich <>
	- add C API stuff. From Joshua Pritikin
	- VMS Makefile.PL fun.	From (Peter Prymmer)
	- hopefully correct "-lc" fix for SCO.
	- add PPD stuff

1.19  Tue Sep 29 22:30 1998
	- put VMS gettimeofday() in. Patch is from Sebastian Bazley
	- change GIMME_V to GIMME to help people with older versions of
	- fix Win32 version of gettimeofday(). It didn't affect anything,
	  but it confuses people reading the code when the return value
	  is backwards (0 is success).
	- fix Makefile.PL (more) so that detection of gettimeofday is
	  more correct.

1.18  Mon Jul 6 22:40 1998
	- add usleep() for Win32.
	- fix Makefile.PL to fix reported HP/UX feature where unresolved
	  externals still cause an executable to be generated (though no
	  x bit set). Thanks to David Kozinn for report and explanation.
	  Problems with the fix are mine :)

1.17  Wed Jul 1 20:10 1998
	- fix setitimer calls so microseconds is not more than 1000000.
	  Hp/UX 9 doesn't like that. Provided by Roland B Robert, PhD.
	- make Win32. We only get gettimeofday (the select hack doesn't
	  seem to work on my Win95 system).
	- fix test 4 on 01test.t. add test to see if time() and 
	  Time::HiRes::time() are close.

1.16  Wed Nov 12 21:05 1997
	- add missing EXTEND in new gettimeofday scalar code.

1.15  Mon Nov 10 21:30 1997
	- update pod. Provided by Gisle Aas.
	- HiRes.xs: if gettimeofday() called in scalar context, do
	  something more useful than before. Provided by Gisle Aas.
	- README: tell of xsubpp '-nolinenumber' woes. thanks to
	  Edward Henigin <> for pointing out the problem.

1.14  Wed Nov 5 9:40 1997
	- Makefile.PL: look for setitimer
	- HiRes.xs: if missing ualarm, but we have setitimer, make up
	  our own setitimer. These were provided by Gisle Aas.

1.13  Tue Nov 4 23:30 1997
	- Makefile.PL: fix autodetect mechanism to do try linking in addition
	  to just compiling; should fix Linux build problem. Fix was provided
	  by Gisle Aas.

1.12  Sun Oct 12 12:00:00 1997
	- Makefile.PL: set XSOPT to '-nolinenumbers' to work around xsubpp bug;
	  you may need to comment this back out if you have an older xsubpp.

1.11  Fri Sep 05 16:00:00 1997
	- Makefile.PL:
	  Had some line commented out that shouldn't have been (testing
	  Previous version was corrupted.

1.10  Thu May 22 20:20:00 1997
	- HiRes.xs,, t/*:
	      -	only compile what we have OS support for (or can 
		fake with select())
	      - only test what we compiled 
	      - gross improvement to the test suite
	      - fix EXPORT_FAIL. 
	  This work was all done by Roderick Schertler
	  <>. If you run Linux or
	  one of the other ualarm-less platoforms, and you like this 
	  module, let Roderick know; without him, it still wouldn't 
	  be working on those boxes...
	- Makefile.PL: figure out what routines the OS has and
	  only build what we need. These bits were written by Jarkko 
	  Hietaniemi <>. Again, gratitude is due...

1.02  Mon Dec 30 08:00:00 1996
	- update documentation to say what to do when missing
	  ualarm() and friends.
	- README: update to warn that ualarm() and friends need to exist

1.01  Fri Oct 17 08:00:00 1996
	- Makefile.PL: make XSPROTOARGS => '-noprototyopes'
	- put blank line between __END__ and =head1 so that 
	  pod2man works.

1.00  Tue Sep 03 13:00:00 1996
	- original version; created by h2xs 1.16