FAIL test/regress_dns.c:1558: assert(be_outcome[3].dnserr != 0): 0 vs 0dns/bufferevent_connect_hostname_emfile: [msg] Nameserver 127.0.0.1:55375 has failed: request timed out.

[psumbera]

gmake check (with latest sources) fails on Solaris with following error:

./test/test.sh -b SELECT
Running tests:
SELECT
 test-eof: OKAY
 test-closed: [warn] event_base_new_with_config: no event mechanism available
OKAY
 test-weof: OKAY
 test-time: OKAY
 test-changelist: OKAY
 test-fdleak: OKAY
 test-dumpevents: OKAY
 regress:
  FAIL test/regress_dns.c:1558: assert(be_outcome[3].dnserr != 0): 0 vs 0dns/bufferevent_connect_hostname_emfile: [msg] Nameserver 127.0.0.1:55375 has failed: request timed out.
[msg] All nameservers have failed
[warn] EVDNS_SOPT_TCP_MAX_CLIENTS option can be set only on TCP server
[warn] EVDNS_SOPT_TCP_MAX_CLIENTS option can be set only on TCP server
[warn] EVDNS_SOPT_TCP_IDLE_TIMEOUT option can be set only on TCP server
[warn] EVDNS_SOPT_TCP_IDLE_TIMEOUT option can be set only on TCP server
1/381 TESTS FAILED. (46 skipped)
FAILED
 regress_debug:
  FAIL test/regress_dns.c:1558: assert(be_outcome[3].dnserr != 0): 0 vs 0dns/bufferevent_connect_hostname_emfile: [msg] Nameserver 127.0.0.1:60405 has failed: request timed out.
[msg] All nameservers have failed
[warn] EVDNS_SOPT_TCP_MAX_CLIENTS option can be set only on TCP server
[warn] EVDNS_SOPT_TCP_MAX_CLIENTS option can be set only on TCP server
[warn] EVDNS_SOPT_TCP_IDLE_TIMEOUT option can be set only on TCP server
[warn] EVDNS_SOPT_TCP_IDLE_TIMEOUT option can be set only on TCP server
1/381 TESTS FAILED. (46 skipped)
FAILED
gmake[3]: *** [Makefile:3447: test_runner_select] Error 1

[azat]

Hello! Thank you for the report. Is this error permanent? There are some well known flakiness actually.

[azat]

There are some well known flakiness actually.

But this is not one of them. Interesting why it does not fail, can you share strace.log and ltrace.log:

• strace -s 1000 -o strace.log regress --no-fork dns/bufferevent_connect_hostname_emfile
• ltrace -s 1000 -o ltrace.log regress --no-fork dns/bufferevent_connect_hostname_emfile

[psumbera]

It happens always when I run: test.sh -b SELECT.

Note that this is Solaris so we don't have strace. Instead we have truss.

But when I run directly regress nothing happens:

 ./regress --no-fork dns/bufferevent_connect_hostname_emfile
dns/bufferevent_connect_hostname_emfile: OK
1 tests ok.  (0 skipped)

And when I run it without test name it fails like this:

./regress --no-fork
main/methods: OK
main/version: OK
..
main/gettimeofday_cached_disabled_nosleep: OK
main/active_by_fd: OK
main/fork: OK
main/del_wait: [warn] Can't change lock callbacks once they have been initialized.
[err] evthread.c:277: Assertion lock->held_by == me failed in evthread_debug_lock_mark_unlocked
Abort (core dumped)

[azat]

Interesting here is completely different error, it got SIGSEGV. If you have an ability to post stacktraces, dtruss output something else, that will be great. Otherwise this will wait until I will setup env with solaris.