Christian Bartolomäus

For the record: The change in PR #893 didn't have a measurable effect when compiling my code base at $work.

@tskirvin Would you mind to re-run your tests with 2.4.2? Maybe the issue can be closed finally :)

After taking a look at the tests and the implementation I read this issue as a feature request. The current implementation doesn't try to warn on unrecognized escape sequences and...

If I'm reading the old examples correctly, the behavior hasn't changed. No error is reported if the variable name *is not* within curly braces -- regardless whether the ```$``` sign...

There is a problem when running with `RAKUDO_RAKUAST=1` with one of the tests: ``` $ RAKULIB=lib RAKUDO_RAKUAST=1 ./rakudo-m t/spec/S04-statements/label.t [...] not ok 9 - no crash: nested loops and labeled...

This does not happen on the JVM backend. Also, the error goes away on MoarVM with ```--optimize=1``` (or below).

If I revert the relevant change from https://github.com/rakudo/rakudo/commit/e2ec1607f9 as follows: ``` diff --git a/src/core.c/Str.pm6 b/src/core.c/Str.pm6 index 363e2dcdf..09d61b163 100644 --- a/src/core.c/Str.pm6 +++ b/src/core.c/Str.pm6 @@ -3579,7 +3579,7 @@ my class Str does...

So a golf of the underlying problem seems to be something like this (works on MoarVM): ``` $ ./rakudo-j -e 'use nqp; my $foo := nqp::box_s("2", IntStr.new(2, "2")); say "alive";...

After looking some more, I'd think this could be the default value for the ```Int``` slot for allomorphs leaking through. This seems to be ```null``` (in this specific case) on...

I can confirm that the test file passes again on the JVM backend. But since this went totally unnoticed for MoarVM, shouldn't we add a test? If not in roast...