| Subject: | Memory cycle (possible leak) with die in async sub |
Test case:
$ cat memleak.pl
#!/usr/bin/perl
use strict;
use warnings;
use Future;
use Future::AsyncAwait;
use Devel::MAT::Dumper;
async sub func
{
await Future->new;
}
eval {
my $f = func();
Devel::MAT::Dumper::dump( "d1.pmat" );
die "Oopsie\n";
};
Devel::MAT::Dumper::dump( "d2.pmat" );
print STDERR "EOF\n";
$ perl memleak.pl
EOF
$ pmat d2.pmat
pmat: find blessed Future
HASH(3)=Future at 0x5618574c72d8: Future
pmat: identify 0x5618574c72d8
HASH(3)=Future at 0x5618574c72d8 is:
└─(via RV) a suspended mortal of CODE(PP) at 0x5618574c72c0, which is:
└─(via RV) element [1] of ARRAY(2) at 0x5618574d1130, which is:
└─(via RV) element [0] of ARRAY(1) at 0x5618574a4680, which is:
└─(via RV) value {callbacks} of HASH(3)=Future at 0x5618574c72d8, which is:
└─already found circularly circularly
Future is being kept alive purely by a suspended mortal from its own tmpstack
--
Paul Evans