From @nanis

When running `nmake spectest` with a development snapshot build of
MoarVM, NQP, and Rakudo, I noticed that the socket test is hanging​:

t\spec\S32-io\IO-Socket-INET.t .................................... 49/51

I have to manually kill the moarvm.exe instance running this
particular test to let `nmake spectest` continue.

Running with higher verbosity gives me the following output *after* I
terminate the process running the test from the Task Manager​:

D​:\Src\perl6\rakudo> perl t\harness5 --verbosity=5
t\spec\S32-io\IO-Socket-INET.t
t\spec\S32-io\IO-Socket-INET.t ..
1..51
# [0s] starting tests
# [0s] Testing on port 1024
ok 1 - [2s] echo server and client
ok 2 - finished in time \#​1
ok 3 - [1s] discard server and client
ok 4 - finished in time \#​2
ok 5 - [2s] received first 7 characters
ok 6 - finished in time \#​3
ok 7 - [0s] received next 3 characters
ok 8 - finished in time \#​4
ok 9 - [0s] remaining 26 were buffered
ok 10 - finished in time \#​5
ok 11 - [0s] received ꀁ
ok 12 - finished in time \#​6
ok 13 - [0s] ... which is 1 character
ok 14 - finished in time \#​7
ok 15 - [0s] received another character
ok 16 - finished in time \#​8
ok 17 - [0s] combined the bytes form ꀂ
ok 18 - finished in time \#​9
ok 19 - [2s] get() with default separator
ok 20 - finished in time \#​10
ok 21 - [0s] default separator
ok 22 - finished in time \#​11
ok 23 - [0s] \r\n separator
ok 24 - [0s] \r was not left behind on the string
ok 25 - finished in time \#​12
ok 26 - [0s] . as a separator
ok 27 - finished in time \#​13
ok 28 - [0s] ! separator not at end of string
ok 29 - finished in time \#​14
ok 30 - [0s] Multiple separators not at end of string
ok 31 - finished in time \#​15
ok 32 - [0s] ! separator at end of string
ok 33 - [5s] received first character
ok 34 - finished in time \#​16
ok 35 - [0s] received last character
ok 36 - finished in time \#​17
ok 37 - [0s] total amount
ok 38 - finished in time \#​18
ok 39 - [1s] successful read binary data
ok 40 - finished in time \#​19
ok 41 - [2s] successful received binary data
ok 42 - finished in time \#​20
ok 43 - [2s] received first character
ok 44 - finished in time \#​21
ok 45 - [0s] received last character
ok 46 - finished in time \#​22
ok 47 - [0s] total amount
ok 48 - finished in time \#​23
ok 49 - [2s] echo server and client
ok 50 - finished in time \#​1
Dubious, test returned 1 (wstat 256, 0x100)
Failed 1/51 subtests

Test Summary Report

t\spec\S32-io\IO-Socket-INET.t (Wstat​: 256 Tests​: 50 Failed​: 0)
  Non-zero exit status​: 1
  Parse errors​: Bad plan. You planned 51 tests but ran 50.
Files=1, Tests=50, 179 wallclock secs ( 0.08 usr + 0.01 sys = 0.09 CPU)
Result​: FAIL

From @zoffixznet

Hi, would you please include the version you're using?. You can get it by running `perl6 -v`

The RT System itself - Status changed from 'new' to 'open'

From @nanis

I have observed this before (including with the 2017.01 release), but
most recently today after building from a fresh pull.

$ perl6 -v
This is Rakudo version 2017.02-140-gdd1dab2 built on MoarVM version
2017.02-20-g773711e
implementing Perl 6.c.

$ nqp -v
This is nqp version 2017.02-39-g2780ed7 built on MoarVM version
2017.02-20-g773711e

On Wed, Mar 1, 2017 at 5​:04 PM, Zoffix Znet via RT
<perl6-bugs-followup@​perl.org> wrote​:

Hi, would you please include the version you're using?. You can get it by running `perl6 -v`

From @ronaldxs

From your description of running the test with nmake I find it reasonable to assume you are testing on windows. Winsock seems to have a 1 second delay for “connection refused” errors which is triggered (2000 times) by a test added with PR 227​: Raku/roast#227. I saw some documentation of the delay on msdn (https://social.msdn.microsoft.com/Forums/sqlserver/en-US/9b90b65c-4630-46b1-98f1-5d4d22427962/delayed-connection-refused-error?forum=netfxnetcom ) and is demonstrated by a small Perl 5 script below. The problem also effects roast on WSL (windows subsystem for linux).

# test below runs in about 10 seconds on windows/WSL
# and instantly on linux

use Socket; # This defines PF_INET and SOCK_STREAM

for (1..10) {
  socket my $socket,PF_INET,SOCK_STREAM,(getprotobyname('tcp'))[2];
  connect $socket, sockaddr_in(1, inet_aton('127.0.0.1'))
  or $! =~ /\bconnection\b.*\brefused\b/i
  or die "can't connect​: $!\n";
  close($socket);
}

From @zoffixznet

A Windows-specific skip-fudge now went in[^1] but the problem remains

[1] Raku/roast@86abb53