summaryrefslogtreecommitdiff
path: root/src/zmq.cpp
blob: b533d23e405270dd2d85204c7c7c06b5359d7887 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
/*
    Copyright (c) 2007-2011 iMatix Corporation
    Copyright (c) 2007-2011 Other contributors as noted in the AUTHORS file

    This file is part of 0MQ.

    0MQ is free software; you can redistribute it and/or modify it under
    the terms of the GNU Lesser General Public License as published by
    the Free Software Foundation; either version 3 of the License, or
    (at your option) any later version.

    0MQ is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU Lesser General Public License for more details.

    You should have received a copy of the GNU Lesser General Public License
    along with this program.  If not, see <http://www.gnu.org/licenses/>.
*/

#include "platform.hpp"

//  On AIX, poll.h has to be included before zmq.h to get consistent
//  definition of pollfd structure (AIX uses 'reqevents' and 'retnevents'
//  instead of 'events' and 'revents' and defines macros to map from POSIX-y
//  names to AIX-specific names).
#if defined ZMQ_HAVE_LINUX || defined ZMQ_HAVE_FREEBSD ||\
    defined ZMQ_HAVE_OPENBSD || defined ZMQ_HAVE_SOLARIS ||\
    defined ZMQ_HAVE_OSX || defined ZMQ_HAVE_QNXNTO ||\
    defined ZMQ_HAVE_HPUX || defined ZMQ_HAVE_AIX ||\
    defined ZMQ_HAVE_NETBSD
#include <poll.h>
#endif

#include "../include/zmq.h"
#include "../include/zmq_utils.h"

#include <string.h>
#include <errno.h>
#include <stdlib.h>
#include <new>

#include "socket_base.hpp"
#include "msg_content.hpp"
#include "stdint.hpp"
#include "config.hpp"
#include "likely.hpp"
#include "clock.hpp"
#include "ctx.hpp"
#include "err.hpp"
#include "fd.hpp"

#if !defined ZMQ_HAVE_WINDOWS
#include <unistd.h>
#endif

#if defined ZMQ_HAVE_OPENPGM
#define __PGM_WININT_H__
#include <pgm/pgm.h>

//  TODO: OpenPGM redefines bool -- remove this once OpenPGM is fixed.
#if defined bool
#undef bool
#endif

#endif

void zmq_version (int *major_, int *minor_, int *patch_)
{
    *major_ = ZMQ_VERSION_MAJOR;
    *minor_ = ZMQ_VERSION_MINOR;
    *patch_ = ZMQ_VERSION_PATCH;
}

const char *zmq_strerror (int errnum_)
{
    return zmq::errno_to_string (errnum_);
}

int zmq_msg_init (zmq_msg_t *msg_)
{
    msg_->content = (zmq::msg_content_t*) ZMQ_VSM;
    msg_->flags = 0;
    msg_->vsm_size = 0;
    return 0;
}

int zmq_msg_init_size (zmq_msg_t *msg_, size_t size_)
{
    if (size_ <= ZMQ_MAX_VSM_SIZE) {
        msg_->content = (zmq::msg_content_t*) ZMQ_VSM;
        msg_->flags = 0;
        msg_->vsm_size = (uint8_t) size_;
    }
    else {
        msg_->content =
            (zmq::msg_content_t*) malloc (sizeof (zmq::msg_content_t) + size_);
        if (!msg_->content) {
            errno = ENOMEM;
            return -1;
        }
        msg_->flags = 0;
        
        zmq::msg_content_t *content = (zmq::msg_content_t*) msg_->content;
        content->data = (void*) (content + 1);
        content->size = size_;
        content->ffn = NULL;
        content->hint = NULL;
        new (&content->refcnt) zmq::atomic_counter_t ();
    }
    return 0;
}

int zmq_msg_init_data (zmq_msg_t *msg_, void *data_, size_t size_,
    zmq_free_fn *ffn_, void *hint_)
{
    msg_->content = (zmq::msg_content_t*) malloc (sizeof (zmq::msg_content_t));
    alloc_assert (msg_->content);
    msg_->flags = 0;
    zmq::msg_content_t *content = (zmq::msg_content_t*) msg_->content;
    content->data = data_;
    content->size = size_;
    content->ffn = ffn_;
    content->hint = hint_;
    new (&content->refcnt) zmq::atomic_counter_t ();
    return 0;
}

int zmq_msg_close (zmq_msg_t *msg_)
{
    //  For VSMs and delimiters there are no resources to free.
    if (msg_->content != (zmq::msg_content_t*) ZMQ_DELIMITER &&
          msg_->content != (zmq::msg_content_t*) ZMQ_VSM) {

        //  If the content is not shared, or if it is shared and the reference.
        //  count has dropped to zero, deallocate it.
        zmq::msg_content_t *content = (zmq::msg_content_t*) msg_->content;
        if (!(msg_->flags & ZMQ_MSG_SHARED) || !content->refcnt.sub (1)) {

            //  We used "placement new" operator to initialize the reference.
            //  counter so we call its destructor now.
            content->refcnt.~atomic_counter_t ();

            if (content->ffn)
                content->ffn (content->data, content->hint);
            free (content);
        }
    }

    //  As a safety measure, let's make the deallocated message look like
    //  an empty message.
    msg_->content = (zmq::msg_content_t*) ZMQ_VSM;
    msg_->flags = 0;
    msg_->vsm_size = 0;

    return 0;
}

int zmq_msg_move (zmq_msg_t *dest_, zmq_msg_t *src_)
{
    zmq_msg_close (dest_);
    *dest_ = *src_;
    zmq_msg_init (src_);
    return 0;
}

int zmq_msg_copy (zmq_msg_t *dest_, zmq_msg_t *src_)
{
    zmq_msg_close (dest_);

    //  VSMs and delimiters require no special handling.
    if (src_->content != (zmq::msg_content_t*) ZMQ_DELIMITER &&
          src_->content != (zmq::msg_content_t*) ZMQ_VSM) {

        //  One reference is added to shared messages. Non-shared messages
        //  are turned into shared messages and reference count is set to 2.
        zmq::msg_content_t *content = (zmq::msg_content_t*) src_->content;
        if (src_->flags & ZMQ_MSG_SHARED)
            content->refcnt.add (1);
        else {
            src_->flags |= ZMQ_MSG_SHARED;
            content->refcnt.set (2);
        }
    }

    *dest_ = *src_;
    return 0;
}

void *zmq_msg_data (zmq_msg_t *msg_)
{
    if (msg_->content == (zmq::msg_content_t*) ZMQ_VSM)
        return msg_->vsm_data;
    if (msg_->content == (zmq::msg_content_t*) ZMQ_DELIMITER)
        return NULL;

    return ((zmq::msg_content_t*) msg_->content)->data;
}

size_t zmq_msg_size (zmq_msg_t *msg_)
{
    if (msg_->content == (zmq::msg_content_t*) ZMQ_VSM)
        return msg_->vsm_size;
    if (msg_->content == (zmq::msg_content_t*) ZMQ_DELIMITER)
        return 0;

    return ((zmq::msg_content_t*) msg_->content)->size;
}

void *zmq_init (int io_threads_)
{
    if (io_threads_ < 0) {
        errno = EINVAL;
        return NULL;
    }

#if defined ZMQ_HAVE_OPENPGM

    //  Init PGM transport. Ensure threading and timer are enabled. Find PGM
    //  protocol ID. Note that if you want to use gettimeofday and sleep for
    //  openPGM timing, set environment variables PGM_TIMER to "GTOD" and
    //  PGM_SLEEP to "USLEEP".
    pgm_error_t *pgm_error = NULL;
    const bool ok = pgm_init (&pgm_error);
    if (ok != TRUE) {

        //  Invalid parameters don't set pgm_error_t
        zmq_assert (pgm_error != NULL);
        if (pgm_error->domain == PGM_ERROR_DOMAIN_TIME && (
              pgm_error->code == PGM_ERROR_FAILED)) {

            //  Failed to access RTC or HPET device.
            pgm_error_free (pgm_error);
            errno = EINVAL;
            return NULL;
        }

        //  PGM_ERROR_DOMAIN_ENGINE: WSAStartup errors or missing WSARecvMsg.
        zmq_assert (false);
    }
#endif

#ifdef ZMQ_HAVE_WINDOWS
    //  Intialise Windows sockets. Note that WSAStartup can be called multiple
    //  times given that WSACleanup will be called for each WSAStartup.
   //  We do this before the ctx constructor since its embedded mailbox_t
   //  object needs Winsock to be up and running.
    WORD version_requested = MAKEWORD (2, 2);
    WSADATA wsa_data;
    int rc = WSAStartup (version_requested, &wsa_data);
    zmq_assert (rc == 0);
    zmq_assert (LOBYTE (wsa_data.wVersion) == 2 &&
        HIBYTE (wsa_data.wVersion) == 2);
#endif

    //  Create 0MQ context.
    zmq::ctx_t *ctx = new (std::nothrow) zmq::ctx_t ((uint32_t) io_threads_);
    alloc_assert (ctx);
    return (void*) ctx;
}

int zmq_term (void *ctx_)
{
    if (!ctx_) {
        errno = EFAULT;
        return -1;
    }

    int rc = ((zmq::ctx_t*) ctx_)->terminate ();
    int en = errno;

#ifdef ZMQ_HAVE_WINDOWS
    //  On Windows, uninitialise socket layer.
    rc = WSACleanup ();
    wsa_assert (rc != SOCKET_ERROR);
#endif

#if defined ZMQ_HAVE_OPENPGM
    //  Shut down the OpenPGM library.
    if (pgm_shutdown () != TRUE)
        zmq_assert (false);
#endif

    errno = en;
    return rc;
}

void *zmq_socket (void *ctx_, int type_)
{
    if (!ctx_) {
        errno = EFAULT;
        return NULL;
    }
    return (void*) (((zmq::ctx_t*) ctx_)->create_socket (type_));
}

int zmq_close (void *s_)
{
    if (!s_) {
        errno = EFAULT;
        return -1;
    }
    ((zmq::socket_base_t*) s_)->close ();
    return 0;
}

int zmq_setsockopt (void *s_, int option_, const void *optval_,
    size_t optvallen_)
{
    if (!s_) {
        errno = EFAULT;
        return -1;
    }
    return (((zmq::socket_base_t*) s_)->setsockopt (option_, optval_,
        optvallen_));
}

int zmq_getsockopt (void *s_, int option_, void *optval_, size_t *optvallen_)
{
    if (!s_) {
        errno = EFAULT;
        return -1;
    }
    return (((zmq::socket_base_t*) s_)->getsockopt (option_, optval_,
        optvallen_));
}

int zmq_bind (void *s_, const char *addr_)
{
    if (!s_) {
        errno = EFAULT;
        return -1;
    }
    return (((zmq::socket_base_t*) s_)->bind (addr_));
}

int zmq_connect (void *s_, const char *addr_)
{
    if (!s_) {
        errno = EFAULT;
        return -1;
    }
    return (((zmq::socket_base_t*) s_)->connect (addr_));
}

int zmq_send (void *s_, const void *buf_, size_t len_, int flags_)
{
    zmq_msg_t msg;
    int rc = zmq_msg_init_size (&msg, len_);
    if (rc != 0)
        return -1;
    memcpy (zmq_msg_data (&msg), buf_, len_);

    rc = zmq_sendmsg (s_, &msg, flags_);
    if (unlikely (rc < 0)) {
        int err = errno;
        int rc2 = zmq_msg_close (&msg);
        errno_assert (rc2 == 0);
        errno = err;
        return -1;
    }
    
    //  Note the optimisation here. We don't close the msg object as it is
    //  empty anyway. This may change when implementation of zmq_msg_t changes.
    return rc;
}

int zmq_recv (void *s_, void *buf_, size_t len_, int flags_)
{
    zmq_msg_t msg;
    int rc = zmq_msg_init (&msg);
    errno_assert (rc == 0);

    rc = zmq_recvmsg (s_, &msg, flags_);
    if (unlikely (rc < 0)) {
        int err = errno;
        int rc2 = zmq_msg_close (&msg);
        errno_assert (rc2 == 0);
        errno = err;
        return -1;
    }

    //  At the moment an oversized message is silently truncated.
    //  TODO: Build in a notification mechanism to report the overflows.
    size_t to_copy = size_t (rc) < len_ ? size_t (rc) : len_;
    memcpy (buf_, zmq_msg_data (&msg), to_copy);
    return (int) to_copy;    
}

int zmq_sendmsg (void *s_, zmq_msg_t *msg_, int flags_)
{
    if (!s_) {
        errno = EFAULT;
        return -1;
    }
    int sz = (int) zmq_msg_size (msg_);
    int rc = (((zmq::socket_base_t*) s_)->send (msg_, flags_));
    if (unlikely (rc < 0))
        return -1;
    return sz;
}

int zmq_recvmsg (void *s_, zmq_msg_t *msg_, int flags_)
{
    if (!s_) {
        errno = EFAULT;
        return -1;
    }
    int rc = (((zmq::socket_base_t*) s_)->recv (msg_, flags_));
    if (unlikely (rc < 0))
        return -1;
    return (int) zmq_msg_size (msg_);
}

#if defined ZMQ_FORCE_SELECT
#define ZMQ_POLL_BASED_ON_SELECT
#elif defined ZMQ_FORCE_POLL
#define ZMQ_POLL_BASED_ON_POLL
#elif defined ZMQ_HAVE_LINUX || defined ZMQ_HAVE_FREEBSD ||\
    defined ZMQ_HAVE_OPENBSD || defined ZMQ_HAVE_SOLARIS ||\
    defined ZMQ_HAVE_OSX || defined ZMQ_HAVE_QNXNTO ||\
    defined ZMQ_HAVE_HPUX || defined ZMQ_HAVE_AIX ||\
    defined ZMQ_HAVE_NETBSD
#define ZMQ_POLL_BASED_ON_POLL
#elif defined ZMQ_HAVE_WINDOWS || defined ZMQ_HAVE_OPENVMS
#define ZMQ_POLL_BASED_ON_SELECT
#endif

int zmq_poll (zmq_pollitem_t *items_, int nitems_, long timeout_)
{
#if defined ZMQ_POLL_BASED_ON_POLL
    if (unlikely (nitems_ < 0)) {
        errno = EINVAL;
        return -1;
    }
    if (unlikely (nitems_ == 0)) {
        if (timeout_ == 0)
            return 0;
#if defined ZMQ_HAVE_WINDOWS
        Sleep (timeout_ > 0 ? timeout_ : INFINITE);
        return 0;
#else
        return usleep (timeout_ * 1000);
#endif
    }

    if (!items_) {
        errno = EFAULT;
        return -1;
    }

    zmq::clock_t clock;
    uint64_t now = 0;
    uint64_t end = 0;

    pollfd *pollfds = (pollfd*) malloc (nitems_ * sizeof (pollfd));
    alloc_assert (pollfds);

    //  Build pollset for poll () system call.
    for (int i = 0; i != nitems_; i++) {

        //  If the poll item is a 0MQ socket, we poll on the file descriptor
        //  retrieved by the ZMQ_FD socket option.
        if (items_ [i].socket) {
            size_t zmq_fd_size = sizeof (zmq::fd_t);
            if (zmq_getsockopt (items_ [i].socket, ZMQ_FD, &pollfds [i].fd,
                &zmq_fd_size) == -1) {
                free (pollfds);
                return -1;
            }
            pollfds [i].events = items_ [i].events ? POLLIN : 0;
        }
        //  Else, the poll item is a raw file descriptor. Just convert the
        //  events to normal POLLIN/POLLOUT for poll ().
        else {
            pollfds [i].fd = items_ [i].fd;
            pollfds [i].events =
                (items_ [i].events & ZMQ_POLLIN ? POLLIN : 0) |
                (items_ [i].events & ZMQ_POLLOUT ? POLLOUT : 0);
        }
    }

    bool first_pass = true;
    int nevents = 0;

    while (true) {

         //  Compute the timeout for the subsequent poll.
         int timeout;
         if (first_pass)
             timeout = 0;
         else if (timeout_ < 0)
             timeout = -1;
         else
             timeout = end - now;

        //  Wait for events.
        while (true) {
            int rc = poll (pollfds, nitems_, timeout);
            if (rc == -1 && errno == EINTR) {
                free (pollfds);
                return -1;
            }
            errno_assert (rc >= 0);
            break;
        }

        //  Check for the events.
        for (int i = 0; i != nitems_; i++) {

            items_ [i].revents = 0;

            //  The poll item is a 0MQ socket. Retrieve pending events
            //  using the ZMQ_EVENTS socket option.
            if (items_ [i].socket) {
                size_t zmq_events_size = sizeof (uint32_t);
                uint32_t zmq_events;
                if (zmq_getsockopt (items_ [i].socket, ZMQ_EVENTS, &zmq_events,
                    &zmq_events_size) == -1) {
                    free (pollfds);
                    return -1;
                }
                if ((items_ [i].events & ZMQ_POLLOUT) &&
                      (zmq_events & ZMQ_POLLOUT))
                    items_ [i].revents |= ZMQ_POLLOUT;
                if ((items_ [i].events & ZMQ_POLLIN) &&
                      (zmq_events & ZMQ_POLLIN))
                    items_ [i].revents |= ZMQ_POLLIN;
            }
            //  Else, the poll item is a raw file descriptor, simply convert
            //  the events to zmq_pollitem_t-style format.
            else {
                if (pollfds [i].revents & POLLIN)
                    items_ [i].revents |= ZMQ_POLLIN;
                if (pollfds [i].revents & POLLOUT)
                    items_ [i].revents |= ZMQ_POLLOUT;
                if (pollfds [i].revents & ~(POLLIN | POLLOUT))
                    items_ [i].revents |= ZMQ_POLLERR;
            }

            if (items_ [i].revents)
                nevents++;
        }

        //  If timout is zero, exit immediately whether there are events or not.
        if (timeout_ == 0)
            break;

        //  If there are events to return, we can exit immediately.
        if (nevents)
            break;

        //  At this point we are meant to wait for events but there are none.
        //  If timeout is infinite we can just loop until we get some events.
        if (timeout_ < 0) {
            if (first_pass)
                first_pass = false;
            continue;
        }

        //  The timeout is finite and there are no events. In the first pass
        //  we get a timestamp of when the polling have begun. (We assume that
        //  first pass have taken negligible time). We also compute the time
        //  when the polling should time out.
        if (first_pass) {
            now = clock.now_ms ();
            end = now + timeout_;
            if (now == end)
                break;
            first_pass = false;
            continue;
        }

        //  Find out whether timeout have expired.
        now = clock.now_ms ();
        if (now >= end)
            break;
    }

    free (pollfds);
    return nevents;

#elif defined ZMQ_POLL_BASED_ON_SELECT

    if (unlikely (nitems_ < 0)) {
        errno = EINVAL;
        return -1;
    }
    if (unlikely (nitems_ == 0)) {
        if (timeout_ == 0)
            return 0;
#if defined ZMQ_HAVE_WINDOWS
        Sleep (timeout_ > 0 ? timeout_ : INFINITE);
        return 0;
#else
        return usleep (timeout_ * 1000);
#endif
    }

    if (!items_) {
        errno = EFAULT;
        return -1;
    }

    zmq::clock_t clock;
    uint64_t now = 0;
    uint64_t end = 0;

    //  Ensure we do not attempt to select () on more than FD_SETSIZE
    //  file descriptors.
    zmq_assert (nitems_ <= FD_SETSIZE);

    fd_set pollset_in;
    FD_ZERO (&pollset_in);
    fd_set pollset_out;
    FD_ZERO (&pollset_out);
    fd_set pollset_err;
    FD_ZERO (&pollset_err);

    zmq::fd_t maxfd = 0;

    //  Build the fd_sets for passing to select ().
    for (int i = 0; i != nitems_; i++) {

        //  If the poll item is a 0MQ socket we are interested in input on the
        //  notification file descriptor retrieved by the ZMQ_FD socket option.
        if (items_ [i].socket) {
            size_t zmq_fd_size = sizeof (zmq::fd_t);
            zmq::fd_t notify_fd;
            if (zmq_getsockopt (items_ [i].socket, ZMQ_FD, &notify_fd,
                &zmq_fd_size) == -1)
                return -1;
            if (items_ [i].events) {
                FD_SET (notify_fd, &pollset_in);
                if (maxfd < notify_fd)
                    maxfd = notify_fd;
            }
        }
        //  Else, the poll item is a raw file descriptor. Convert the poll item
        //  events to the appropriate fd_sets.
        else {
            if (items_ [i].events & ZMQ_POLLIN)
                FD_SET (items_ [i].fd, &pollset_in);
            if (items_ [i].events & ZMQ_POLLOUT)
                FD_SET (items_ [i].fd, &pollset_out);
            if (items_ [i].events & ZMQ_POLLERR)
                FD_SET (items_ [i].fd, &pollset_err);
            if (maxfd < items_ [i].fd)
                maxfd = items_ [i].fd;
        }
    }

    bool first_pass = true;
    int nevents = 0;
    fd_set inset, outset, errset;

    while (true) {

        //  Compute the timeout for the subsequent poll.
        timeval timeout;
        timeval *ptimeout;
        if (first_pass) {
            timeout.tv_sec = 0;
            timeout.tv_usec = 0;
            ptimeout = &timeout;
        }
        else if (timeout_ < 0)
            ptimeout = NULL;
        else {
            timeout.tv_sec = (long) ((end - now) / 1000);
            timeout.tv_usec = (long) ((end - now) % 1000 * 1000);
            ptimeout = &timeout;
        }

        //  Wait for events. Ignore interrupts if there's infinite timeout.
        while (true) {
            memcpy (&inset, &pollset_in, sizeof (fd_set));
            memcpy (&outset, &pollset_out, sizeof (fd_set));
            memcpy (&errset, &pollset_err, sizeof (fd_set));
            int rc = select (maxfd + 1, &inset, &outset, &errset, ptimeout);
#if defined ZMQ_HAVE_WINDOWS
            wsa_assert (rc != SOCKET_ERROR);
#else
            if (rc == -1 && errno == EINTR)
                return -1;
            errno_assert (rc >= 0);
#endif
            break;
        }

        //  Check for the events.
        for (int i = 0; i != nitems_; i++) {

            items_ [i].revents = 0;

            //  The poll item is a 0MQ socket. Retrieve pending events
            //  using the ZMQ_EVENTS socket option.
            if (items_ [i].socket) {
                size_t zmq_events_size = sizeof (uint32_t);
                uint32_t zmq_events;
                if (zmq_getsockopt (items_ [i].socket, ZMQ_EVENTS, &zmq_events,
                      &zmq_events_size) == -1)
                    return -1;
                if ((items_ [i].events & ZMQ_POLLOUT) &&
                      (zmq_events & ZMQ_POLLOUT))
                    items_ [i].revents |= ZMQ_POLLOUT;
                if ((items_ [i].events & ZMQ_POLLIN) &&
                      (zmq_events & ZMQ_POLLIN))
                    items_ [i].revents |= ZMQ_POLLIN;
            }
            //  Else, the poll item is a raw file descriptor, simply convert
            //  the events to zmq_pollitem_t-style format.
            else {
                if (FD_ISSET (items_ [i].fd, &inset))
                    items_ [i].revents |= ZMQ_POLLIN;
                if (FD_ISSET (items_ [i].fd, &outset))
                    items_ [i].revents |= ZMQ_POLLOUT;
                if (FD_ISSET (items_ [i].fd, &errset))
                    items_ [i].revents |= ZMQ_POLLERR;
            }

            if (items_ [i].revents)
                nevents++;
        }

        //  If timout is zero, exit immediately whether there are events or not.
        if (timeout_ == 0)
            break;

        //  If there are events to return, we can exit immediately.
        if (nevents)
            break;

        //  At this point we are meant to wait for events but there are none.
        //  If timeout is infinite we can just loop until we get some events.
        if (timeout_ < 0) {
            if (first_pass)
                first_pass = false;
            continue;
        }

        //  The timeout is finite and there are no events. In the first pass
        //  we get a timestamp of when the polling have begun. (We assume that
        //  first pass have taken negligible time). We also compute the time
        //  when the polling should time out.
        if (first_pass) {
            now = clock.now_ms ();
            end = now + timeout_;
            if (now == end)
                break;
            first_pass = false;
            continue;
        }

        //  Find out whether timeout have expired.
        now = clock.now_ms ();
        if (now >= end)
            break;
    }

    return nevents;

#else
    //  Exotic platforms that support neither poll() nor select().
    errno = ENOTSUP;
    return -1;
#endif
}

#if defined ZMQ_POLL_BASED_ON_SELECT
#undef ZMQ_POLL_BASED_ON_SELECT
#endif
#if defined ZMQ_POLL_BASED_ON_POLL
#undef ZMQ_POLL_BASED_ON_POLL
#endif

int zmq_errno ()
{
    return errno;
}

////////////////////////////////////////////////////////////////////////////////
//  0MQ utils - to be used by perf tests
////////////////////////////////////////////////////////////////////////////////

void zmq_sleep (int seconds_)
{
#if defined ZMQ_HAVE_WINDOWS
    Sleep (seconds_ * 1000);
#else
    sleep (seconds_);
#endif
}

void *zmq_stopwatch_start ()
{
    uint64_t *watch = (uint64_t*) malloc (sizeof (uint64_t));
    alloc_assert (watch);
    *watch = zmq::clock_t::now_us ();
    return (void*) watch;
}

unsigned long zmq_stopwatch_stop (void *watch_)
{
    uint64_t end = zmq::clock_t::now_us ();
    uint64_t start = *(uint64_t*) watch_;
    free (watch_);
    return (unsigned long) (end - start);
}