ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/EV/Makefile.PL
(Generate patch)

Comparing EV/Makefile.PL (file contents):
Revision 1.38 by root, Thu Nov 20 00:35:16 2008 UTC vs.
Revision 1.45 by root, Fri Oct 29 19:12:19 2010 UTC

1use 5.006; 1use 5.008002;
2 2
3use strict qw(vars subs); 3use strict qw(vars subs);
4use Config; 4use Config;
5use ExtUtils::MakeMaker; 5use ExtUtils::MakeMaker;
6 6
… …
35 35
36*** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** 36*** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***
37 37
38 38
39POSIX optionally offers support for a monotonic clock source. EV 39POSIX optionally offers support for a monotonic clock source. EV
40can take advantage of this clock source to detect time jumps 40can take advantage of this clock source to detect time jumps more
41reliably. Unfortunately, some systems are bound to be broken, so you can 41reliably. Unfortunately, some systems are bound to be broken, so you can
42disable this here: you can completely disable the detection and use of 42disable this here: you can completely disable the detection and use of
43the monotonic clock by answering 'n' here. Support for this clock type 43the monotonic clock by answering 'n' here. Support for this clock type
44will otherwise be autodetected at both compile- and runtime. (this setting 44will otherwise be autodetected at both compile- and runtime. (this setting
45currently affects the use of nanosleep over select as well). 45currently affects the use of nanosleep over select as well).
… …
54 54
55 55
56POSIX optionally offers support for a (potentially) high-resolution 56POSIX optionally offers support for a (potentially) high-resolution
57realtime clock interface. In a good implementation, using it is faster 57realtime clock interface. In a good implementation, using it is faster
58than the normal method of using gettimeofday. Unfortunately, this option 58than the normal method of using gettimeofday. Unfortunately, this option
59is also bound to be broken on some systems, so you can disable use and 59is also bound to be broken on some systems, and current EV versions do not
60probing of this feature altogether here. Otherwise support for this clock 60actually call gettimeofday very often, so it defaults to no.
61type will be autodetected at compiletime.
62 61
63EOF 62EOF
64 63
65$DEFINE .= " -DEV_USE_REALTIME=" . (0 + (prompt ("Prefer clock_gettime (CLOCK_REALTIME) over gettimeofday (y/n)?", "y") =~ /[yY]/)); 64$DEFINE .= " -DEV_USE_REALTIME=" . (0 + (prompt ("Prefer clock_gettime (CLOCK_REALTIME) over gettimeofday (y/n)?", "n") =~ /[yY]/));
66 65
67print <<EOF; 66print <<EOF;
68 67
69*** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** 68*** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***
70 69
… …
143many BSD systems. Support for kqueue will be detected at runtime, with a 142many BSD systems. Support for kqueue will be detected at runtime, with a
144safe fallback to other methods when it cannot be used. 143safe fallback to other methods when it cannot be used.
145 144
146Note that kqueue is broken on most operating systems, so by default it 145Note that kqueue is broken on most operating systems, so by default it
147won't be used on many platforms, but you can still create your own event 146won't be used on many platforms, but you can still create your own event
148loop with kqueue backend. 147loop with kqueue backend if you ask specifically for it.
149 148
150Here is what we know: 149Here is what we know:
151 150
152NetBSD: partially working in at least 3.1. Yeah! :) 151NetBSD: partially working in at least 3.1 and later. Yeah! :)
153FreeBSD: broken on at least 6.2-STABLE, 152FreeBSD: broken on at least 6.2-STABLE, spotty in later versions,
154 sockets *likely* work, ptys definitely don't. 153 sockets *likely* work, ptys definitely don't.
155OpenBSD: reports indicate that it likely doesn't work 154OpenBSD: reports indicate that it likely doesn't work
156 (similar problems as on FreeBSD). 155 (similar problems as on FreeBSD).
157OS X: completely, utterly broken on at least < 10.5. 156OS X: completely, utterly broken on at least <= 10.6.
158 157
159EOF 158EOF
160 159
161my $can_kqueue = -e "/usr/include/sys/event.h"; 160my $can_kqueue = -e "/usr/include/sys/event.h";
162$can_kqueue = $ENV{EV_KQUEUE} if exists $ENV{EV_KQUEUE}; 161$can_kqueue = $ENV{EV_KQUEUE} if exists $ENV{EV_KQUEUE};
… …
197*** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** 196*** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***
198 197
199 198
200A backend of a different kind is the Linux inotify(7) interface, which can 199A backend of a different kind is the Linux inotify(7) interface, which can
201be used to speed up (and reduce resource consumption) of stat watchers. If 200be used to speed up (and reduce resource consumption) of stat watchers. If
202you have it, it is usually a good idea to enable it. 201you have the include file and libc support for it, it is usually a good
202idea to enable it, as kernel availability is detected at runtime.
203 203
204EOF 204EOF
205 205
206my $can_inotify = -e "/usr/include/sys/inotify.h"; 206my $can_inotify = -e "/usr/include/sys/inotify.h";
207$can_inotify = $ENV{EV_INOTIFY} if exists $ENV{EV_INOTIFY}; 207$can_inotify = $ENV{EV_INOTIFY} if exists $ENV{EV_INOTIFY};
… …
221EOF 221EOF
222 222
223my $can_eventfd = -e "/usr/include/sys/eventfd.h"; 223my $can_eventfd = -e "/usr/include/sys/eventfd.h";
224$can_eventfd = $ENV{EV_EVENTFD} if exists $ENV{EV_EVENTFD}; 224$can_eventfd = $ENV{EV_EVENTFD} if exists $ENV{EV_EVENTFD};
225$DEFINE .= " -DEV_USE_EVENTFD=" . (0 + (prompt ("Enable linux eventfd support (y/n)?", $can_eventfd ? "y" : "n") =~ /[yY]/)); 225$DEFINE .= " -DEV_USE_EVENTFD=" . (0 + (prompt ("Enable linux eventfd support (y/n)?", $can_eventfd ? "y" : "n") =~ /[yY]/));
226
227print <<EOF;
228
229*** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***
230
231
232Another sometimes useful bit of functionality is the Linux signalfd, which
233is useful for faster signal handling (don't care). Kernel support for
234this will be probed at runtime, but your libc must contain the necessary
235wrapper. Glibc 2.7 and later should have this wrapper.
236
237EOF
238
239my $can_signalfd = -e "/usr/include/sys/signalfd.h";
240$can_signalfd = $ENV{EV_SIGNALFD} if exists $ENV{EV_SIGNALFD};
241$DEFINE .= " -DEV_USE_SIGNALFD=" . (0 + (prompt ("Enable linux signalfd support (y/n)?", $can_signalfd ? "y" : "n") =~ /[yY]/));
226 242
227print <<EOF; 243print <<EOF;
228 244
229*** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** 245*** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***
230 246
… …
248 DEFINE => "$DEFINE", 264 DEFINE => "$DEFINE",
249 NAME => "EV", 265 NAME => "EV",
250 LIBS => [$LIBS], 266 LIBS => [$LIBS],
251 PREREQ_PM => { 267 PREREQ_PM => {
252 @anyevent, 268 @anyevent,
269 "common::sense" => 0,
253 }, 270 },
254 VERSION_FROM => "EV.pm", 271 VERSION_FROM => "EV.pm",
255 PM => { 272 PM => {
256 'EV.pm' => '$(INST_LIBDIR)/EV.pm', 273 'EV.pm' => '$(INST_LIBDIR)/EV.pm',
257 'EV/EVAPI.h' => '$(INST_LIBDIR)/EV/EVAPI.h', 274 'EV/EVAPI.h' => '$(INST_LIBDIR)/EV/EVAPI.h',

Diff Legend

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