inconsistent behavior due to my in false conditional

[vividsnow]

Hello

use strict;
use warnings;

($,, $\) = (',', "\n");

# case 1
print 'case 1';
for (1..3) {
    my @v;
    push @v, $_ if 0;
    push @v, $_;
    print @v
}

# case 2
print 'case 2';
for (1..3) {
    push my @v, $_ if 0;
    push @v, $_;
    print @v
}

results:

case 1
1
2
3
case 2
1
1,2
1,2,3

I'm expecting similar output in case 2 as in case 1, but its not and none warnings reported.

Similar issue with other perl versions: https://perlbanjo.com/681bc605bc

[atoomic]

We can have a better idea of what's understood by splitting the two cases in individual programs and inspect the optree

# optree for case 1
> perl -MO=Concise c1.pl
n  <@> leave[1 ref] vKP/REFC ->(end)
1     <0> enter ->2
2     <;> nextstate(main 1 c1.pl:1) v:{ ->3
m     <2> leaveloop vKP/2 ->n
7        <{> enteriter(next->j last->m redo->8) vKS/DEF ->k
-           <0> ex-pushmark s ->3
-           <1> ex-list lK ->6
3              <0> pushmark s ->4
4              <$> const(IV 1) s ->5
5              <$> const(IV 3) s ->6
6           <$> gv(*_) s ->7
-        <1> null vK/1 ->m
l           <|> and(other->8) vK/1 ->m
k              <0> iter s ->l
-              <@> lineseq vKP ->-
8                 <;> nextstate(main 4 c1.pl:2) v ->9
9                 <0> padav[@v:4,5] vM/LVINTRO ->a
-                 <;> ex-nextstate(main 5 c1.pl:3) v ->-
-                 <0> ex-const v/SHORT ->a
a                 <;> nextstate(main 5 c1.pl:4) v:{ ->b
e                 <@> push[t4] vK/2 ->f
b                    <0> pushmark s ->c
c                    <0> padav[@v:4,5] lRM ->d
-                    <1> ex-rv2sv sK/1 ->e
d                       <$> gvsv(*_) s ->e
f                 <;> nextstate(main 5 c1.pl:5) v:{ ->g
i                 <@> print vK ->j
g                    <0> pushmark s ->h
h                    <0> padav[@v:4,5] l ->i
j                 <0> unstack v ->k
c1.pl syntax OK

# optree for case 2
> perl -MO=Concise c2.pl
n  <@> leave[1 ref] vKP/REFC ->(end)
1     <0> enter ->2
2     <;> nextstate(main 1 c2.pl:1) v:{ ->3
m     <2> leaveloop vKP/2 ->n
7        <{> enteriter(next->j last->m redo->8) vKS/DEF ->k
-           <0> ex-pushmark s ->3
-           <1> ex-list lK ->6
3              <0> pushmark s ->4
4              <$> const(IV 1) s ->5
5              <$> const(IV 3) s ->6
6           <$> gv(*_) s ->7
-        <1> null vK/1 ->m
l           <|> and(other->a) vK/1 ->m
k              <0> iter s ->l
-              <@> lineseq vKP ->-
8                 <;> ex-nextstate(main 4 c2.pl:2) v ->9
9                 <0> ex-const v/SHORT ->a
a                 <;> nextstate(main 5 c2.pl:3) v ->b
e                 <@> push[t4] vK/2 ->f
b                    <0> pushmark s ->c
c                    <0> padav[@v:4,5] lRM ->d
-                    <1> ex-rv2sv sK/1 ->e
d                       <$> gvsv(*_) s ->e
f                 <;> nextstate(main 5 c2.pl:4) v:{ ->g
i                 <@> print vK ->j
g                    <0> pushmark s ->h
h                    <0> padav[@v:4,5] l ->i
j                 <0> unstack v ->k
c2.pl syntax OK

The main difference is the extra padav in case 1.
9 <0> padav[@v:4,5] vM/LVINTRO ->a

IMO this is a bug in your code and I would say that the behavior is the expected one.
I think the second case should raise a warning.

[Grinnz]

The second case is just another example of a bug from a "my" in false conditional, as discussed in the tickets linked here: https://rt-archive.perl.org/perl5/Ticket/Display.html?id=89652

[atoomic]

view #11293

[Grinnz]

Now represented by the label: https://github.com/Perl/perl5/labels/meta-my-in-false-conditional

[vividsnow]

IMO this is a bug in your code and I would say that the behavior is the expected one.
I think the second case should raise a warning.

I agree that either both cases should work similarly or case 2 raises error under strict pragma