Signal::StackTrace not working?

perlancar has asked for the wisdom of the Perl Monks concerning the following question:

Just want to check with everybody, is it just me or are Signal::StackTrace and its fork Signal::StackTrace::CarpLike not working properly? At least on Linux (I'm using 5.18.1).

#!perl
use Signal::StackTrace qw(INT);

sub f { g() }
sub g { sleep 10 }

f;
[download]

If I press Ctrl-C in the middle of run, the program above should print a stack trace with f, g included in the trace levels. Which will happen if I replace "sleep" with "die". Instead, what I'm getting is just:

Caller level 0:
{
  Bitmask => 'UUUUUUUUUUUUUU',
  Evaltext => undef,
  Filename => '/path/to/script',
  Hasargs => 1,
  Hints => 0,
  'Line-No' => 0,
  Package => 'main',
  Require => undef,
  Subroutine => 'Signal::StackTrace::__ANON__',
  Wantarray => ''
}
Caller level 1:
{
  Bitmask => 'UUUUUUUUUUUUUU',
  Evaltext => undef,
  Filename => '/path/to/script',
  Hasargs => 0,
  Hints => 2018,
  'Line-No' => 0,
  Package => 'main',
  Require => undef,
  Subroutine => '(eval)',
  Wantarray => ''
}

Comment on Signal::StackTrace not working? Download Code

Replies are listed 'Best First'.

Re: Signal::StackTrace not working?
by Athanasius (Archbishop) on Oct 12, 2014 at 03:13 UTC

Hello perlancar,

I’m seeing the same behaviour on Windows. I think the problem is that the trace is given from the vantage point of the statement immediately following the statement that was in effect when the signal was received. I played around with the following script:

#!perl
use strict;
use warnings;
use Signal::StackTrace::CarpLike qw( INT );

f();
print "After f()\n";

sub f { g(); }
sub g
{
    sleep 5;
    print "In g(), after sleep\n";
}
[download]

Output:

13:05 >perl 1052_SoPW.pl
INT at 1052_SoPW.pl line 13.
        eval {...} called at 1052_SoPW.pl line 13
        main::g() called at 1052_SoPW.pl line 9
        main::f() called at 1052_SoPW.pl line 6
In g(), after sleep
After f()

13:05 >
[download]

Note the line number: line 13 is the print statement following sleep. And with that print statement commented out:

13:05 >perl 1052_SoPW.pl
INT at 1052_SoPW.pl line 7.
        eval {...} called at 1052_SoPW.pl line 7
After f()

13:07 >
[download]

The INT signal is now reported as having been received after sub g returned, because sleep happened to be the last statement in the sub. And the stack trace shown is correct as from the viewpoint of line 7.

This looks like a fairly serious bug to me.

Hope that helps,

Athanasius <°(((>< contra mundum Iustus alius egestas vitae, eros Piratica,

[reply]
[d/l]
[select]