From zefram@fysh.org

When shifting a non-zero integer left by a very large amount, generally
some kind of error should result, such as memory exhaustion. But for
some values Rakudo is producing silly small outputs​:

123 +< (1 +< 31)
8863084066665136128
123 +< ((1 +< 31) + 1)
17726168133330272256
123 +< ((1 +< 32) - 1)
0
123 +< (1 +< 32)
123
123 +< ((1 +< 32) + 1)
246
123 +< ((1 +< 33) - 1)
0
123 +< (1 +< 33)
123
123 +< ((1 +< 33) + 1)
246
123 +< ((1 +< 64) - 1)
61

-zefram

From @lizmat

On 17 Dec 2015, at 03​:48, Zefram (via RT) <perl6-bugs-followup@​perl.org> wrote​:

# New Ticket Created by Zefram
# Please include the string​: [perl #​126943]
# in the subject line of all future correspondence about this issue.
# <URL​: https://rt-archive.perl.org/perl6/Ticket/Display.html?id=126943 >

When shifting a non-zero integer left by a very large amount, generally
some kind of error should result, such as memory exhaustion. But for
some values Rakudo is producing silly small outputs​:

123 +< (1 +< 31)
8863084066665136128
123 +< ((1 +< 31) + 1)
17726168133330272256
123 +< ((1 +< 32) - 1)
0
123 +< (1 +< 32)
123
123 +< ((1 +< 32) + 1)
246
123 +< ((1 +< 33) - 1)
0
123 +< (1 +< 33)
123
123 +< ((1 +< 33) + 1)
246
123 +< ((1 +< 64) - 1)
61

As with #​126941, commit f6091476486d29c8886d gives this a slightly better error message, at least until 6.c.

$ 6 'say 123 +< ((1 +< 32) - 1)'
Big integer 4294967295 not yet supported on righthand side of '+<' operator

Liz

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

From zefram@fysh.org

Elizabeth Mattijsen via RT wrote​:

As with #​126941, commit f6091476486d29c8886d gives this a slightly
better error message, at least until 6.c.

Providing any error here, rather than a wrong answer, resolves this issue.
At least it does so for the cases with a non-zero lhs, for which the
correct answer can't be represented.

In the exception where the lhs is zero, formerly shift distances up to
((1 +< 64) - 1) were accepted, and it's a shame to reduce that range.
But either way that range being limited is LTA in a manner exactly akin
to [perl #​126941]. That ought to be addressed as an extension of that
ticket or as a separate issue.

Since you've raised the question of pre-xmas vs post-xmas, I think the
only one of these bit-shift tickets that could possibly be a pre-xmas
priority is [perl #​126942]. Not because it has any deep implications,
but just because the lower end of its misbehaviour is reachable with
reasonable inputs that are fairly likely to arise in substantive programs.

-zefram