load-yaml not threadsafe

[jnthn]

We've had a test or two that occasionally fails in the cro distribution's test suite. Today I finally got around to hunting it down, and it boiled down to the load-yaml sometimes blowing up if called from two threads at the same time. Thankfully, it's easy to reproduce reliably with a one-liner that users this module (note the second example that crashes adds race):

$ perl6 -e 'use YAMLish; my $yaml = "foo:\n  -bar: 42"; for ^1000 { load-yaml $yaml }'
$ perl6 -e 'use YAMLish; my $yaml = "foo:\n  -bar: 42"; race for ^1000 { load-yaml $yaml }'
Use of Nil in string context
  in block  at /home/jnthn/dev/MoarVM/install/share/perl6/site/sources/9D54CB20630A6D95A459E8D21558CE9627DA4464 (YAMLish) line 145
Use of Nil in string context
  in block  at /home/jnthn/dev/MoarVM/install/share/perl6/site/sources/9D54CB20630A6D95A459E8D21558CE9627DA4464 (YAMLish) line 145
Use of Nil in string context
  in block  at /home/jnthn/dev/MoarVM/install/share/perl6/site/sources/9D54CB20630A6D95A459E8D21558CE9627DA4464 (YAMLish) line 145
A worker in a parallel iteration (hyper or race) initiated here:
  in block <unit> at -e line 1

Died at:
    Couldn't parse YAML
      in sub load-yaml at /home/jnthn/dev/MoarVM/install/share/perl6/site/sources/9D54CB20630A6D95A459E8D21558CE9627DA4464 (YAMLish) line 735
      in block  at -e line 1

[jnthn]

After a look through the module, it's not immediately obvious to me that it's doing anything wrong; it may be a compilation bug with code blocks inside of a before.

[Leont]

My installed perl6 is somewhat old (2017.08-79-g4b02b8aad), I can't reproduce it there. This may be a regression.

[Leont]

Is this still an issue?

[Leont]

The Grammar reenters the Grammar engine at some places, I would guess that to be the root of the issue.

[jnthn]

I think it may be a problem with before; I'll have to dig a bit further.

[japhb]

Behavior recreated with current(ish) Rakudo; line numbers have changed though (and again, no errors without race):

$ perl6 -v
This is Rakudo version 2020.02.1-183-g70ac9b72e built on MoarVM version 2020.02.1-44-g04005cf43
implementing Raku 6.d.

$ perl6 -e 'use YAMLish; my $yaml = "foo:\n -bar: 42"; race for ^1000 { load-yaml $yaml }'
Use of Nil in string context
in block at .../site/sources/ADB31520994572F18CE4EC4457E5E9321A31301C (YAMLish) line 296
Use of Nil in string context
in block at .../site/sources/ADB31520994572F18CE4EC4457E5E9321A31301C (YAMLish) line 289
Use of Nil in string context
in block at .../site/sources/ADB31520994572F18CE4EC4457E5E9321A31301C (YAMLish) line 289
Use of Nil in string context
in block at .../site/sources/ADB31520994572F18CE4EC4457E5E9321A31301C (YAMLish) line 289
A worker in a parallel iteration (hyper or race) initiated here:
in block at -e line 1

Died at:
Couldn't parse YAML
in sub load-yaml at .../site/sources/ADB31520994572F18CE4EC4457E5E9321A31301C (YAMLish) line 951
in block at -e line 1

[pjdurai]

Came here to report this. Can confirm this is still an issue.
I had to put the usage of 'load-yaml' method inside a 'lock.protect' to get my script working.

❯❯❯ rakudo --version
This is Rakudo version 2020.05.1 built on MoarVM version 2020.05
implementing Raku 6.d.

[Leont]

I had to put the usage of 'load-yaml' method inside a 'lock.protect' to get my script working.

Unless @jnthn has a better idea, I guess I'll do that

[patrickbkr]

I am still hitting this with Rakudo 2022.07.

[patrickbkr]

@lizmat Given this is actually a compile time bug of code blocks in before, do I assume correctly, that this issue is likely to be gone on RakuAST all by itself?