From @Ovid

I was trying to write up a new concurrency examples for a talk when I encountered this​:
05​:20​:34 ~/temp $ time perl6 -MHTTP​::UserAgent -e 'my $url = "http://www.rakudo.org/"; my $ua = HTTP​::UserAgent.new; $ua.get($url)'

real 0m12.382s
user 0m10.374s
sys 0m0.203s
05​:21​:33 ~/temp $ time perl6 -MHTTP​::UserAgent -e 'my $url = "http://www.rakudo.org/"; my $ua = HTTP​::UserAgent.new; $ua.get($url); await start { $ua.get($url) }'

real 0m13.869s
user 0m10.607s
sys 0m0.215s
05​:21​:59 ~/temp $ time perl6 -MHTTP​::UserAgent -e 'my $url = "http://www.rakudo.org/"; my $ua = HTTP​::UserAgent.new; await start { $ua.get($url) }'

# hangs forever unless we do a FIRST $ua.get outside of start {}
The last example just hangs. As far as I can tell, I have to make at least one get($url) call outside of the promise for it to work.
I get the same behavior with LWP​::Simple.
05​:25​:35 ~/temp $ time perl6 -MLWP​::Simple -e 'my $url = "http://www.rakudo.org/"; LWP​::Simple.get($url)'
real 0m9.003suser 0m7.277ssys 0m0.145s05​:28​:50 ~/temp $ time perl6 -MLWP​::Simple -e 'my $url = "http://www.rakudo.org/"; LWP​::Simple.get($url); await start { LWP​::Simple.get($url) }'
real 0m11.003suser 0m7.625ssys 0m0.162s05​:29​:11 ~/temp $ time perl6 -MLWP​::Simple -e 'my $url = "http://www.rakudo.org/"; await start { LWP​::Simple.get($url) }'
# hangs ...
Best,Ovid-- IT consulting, training, specializing in Perl, databases, and agile developmenthttp://www.allaroundtheworld.fr/. 
Buy my book! - http://bit.ly/beginning_perl

From ovid@cpan.org

I forgot to mention this is on OS X Yosemite (10.10.5).

This is perl6 version 2015.09-412-g32d7b08 built on MoarVM version 2015.09-79-gee9fc2b

Cheers,
Ovid

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

From @jonathanstowe

I can confirm this happens on other platforms than OS X.

It is not isolated to HTTP​::UserAgent and happens with LWP​::Simple too.

In both cases according to strace it gets hung up in a futex().

I'm struggling to replicate with bare IO​::Socket​::INET code though.

From @jonathanstowe

And the reason that it couldn't be replicated in IO​::Socket​::INET code is that it isn't the socket apparently​:

<ollej> I'm having an issue when using the URI module inside a Promise. Unless I've used the URI module before awaiting the result, the program never finishes. Using rakudo 2015.09.
<ollej> I've boiled it down to the following example​: await Promise.in(1).then({ say URI.new('http://www.example.com'); });

Which can be reduced to​:

use URI;

await start { say URI.new("http://localhost") };

So IO​::Socket​::INET is off the hook.

On Mon Oct 19 14​:41​:59 2015, jns+bc@​gellyfish.co.uk wrote​:

I can confirm this happens on other platforms than OS X.

It is not isolated to HTTP​::UserAgent and happens with LWP​::Simple too.

In both cases according to strace it gets hung up in a futex().

I'm struggling to replicate with bare IO​::Socket​::INET code though.

From @jonathanstowe

Infact​:

<llfourn> perl6 -MIETF​::RFC_Grammar​::URI -e "await Promise.in(1).then({ require IETF​::RFC_Grammar​::URI });" will hang

On Wed Oct 21 03​:33​:33 2015, jns+bc@​gellyfish.co.uk wrote​:

And the reason that it couldn't be replicated in IO​::Socket​::INET code
is that it isn't the socket apparently​:

<ollej> I'm having an issue when using the URI module inside a
Promise. Unless I've used the URI module before awaiting the result,
the program never finishes. Using rakudo 2015.09.
<ollej> I've boiled it down to the following example​: await
Promise.in(1).then({ say URI.new('http://www.example.com'); });

Which can be reduced to​:

use URI;

await start { say URI.new("http://localhost") };

So IO​::Socket​::INET is off the hook.

On Mon Oct 19 14​:41​:59 2015, jns+bc@​gellyfish.co.uk wrote​:

I can confirm this happens on other platforms than OS X.

It is not isolated to HTTP​::UserAgent and happens with LWP​::Simple
too.

In both cases according to strace it gets hung up in a futex().

I'm struggling to replicate with bare IO​::Socket​::INET code though.

From @zoffixznet

To give an update on this. With Rakudo built on Dec 30, 2015, I no longer get any hangs with neither
perl6 -MURI -e 'await start { say URI.new("http://localhost") };'

nor perl6 -MIETF​::RFC_Grammar​::URI -e "await Promise.in(1).then({ require IETF​::RFC_Grammar​::URI });"

This is on Debian Wheezy. I tried both on 32bit 2-core and 64bit 1-core boxes.

From @dogbert17

On Wed, 30 Dec 2015 09​:12​:25 -0800, cpan@​zoffix.com wrote​:

To give an update on this. With Rakudo built on Dec 30, 2015, I no
longer get any hangs with neither
perl6 -MURI -e 'await start { say URI.new("http://localhost") };'

nor perl6 -MIETF​::RFC_Grammar​::URI -e "await Promise.in(1).then({
require IETF​::RFC_Grammar​::URI });"

This is on Debian Wheezy. I tried both on 32bit 2-core and 64bit 1-
core boxes.

This problem has been fixed. Part of the problem seems to have been the
use of require in a thread, i.e. RT #​126587.

@dogbert17 - Status changed from 'open' to 'resolved'

From @AlexDaniel

FWIW it's not possible to reproduce the issue on older rakudos because it was sweeped under the rug in the URI module itself​: raku-community-modules/URI#21

So yes, #​126587 is a “testneeded” ticket for the underlying issue.

On 2018-03-09 12​:58​:27, jan-olof.hendig@​bredband.net wrote​:

On Wed, 30 Dec 2015 09​:12​:25 -0800, cpan@​zoffix.com wrote​:

To give an update on this. With Rakudo built on Dec 30, 2015, I no
longer get any hangs with neither
perl6 -MURI -e 'await start { say URI.new("http://localhost") };'

nor perl6 -MIETF​::RFC_Grammar​::URI -e "await Promise.in(1).then({
require IETF​::RFC_Grammar​::URI });"

This is on Debian Wheezy. I tried both on 32bit 2-core and 64bit 1-
core boxes.

This problem has been fixed. Part of the problem seems to have been the
use of require in a thread, i.e. RT #​126587.