Re^2: Request help for work around for bug in perl 5.10.1 (weak magic)

So I apologize for the lack of clear error messages. The code is detached, so I have to trap __DIE__ and write to syslog. This is done as follows:

$SIG{__DIE__} = \&die_catcher;
...
#
#   die_catcher
#
#   Try to log the call stack if we die
#
sub die_catcher {
    for my $caller ( CallStack() ) {
        alog(LOG_ERR, $caller);
    }
}
...
#
#   Courtesy of Perl Monks http://www.perlmonks.org/?node_id=51097
#
sub CallStack {
    local $@;
    eval { confess( '' ) };
    my @stack = split m/\n/, $@;
    shift @stack for 1..3; # Cover our tracks.
    return wantarray ? @stack : join "\n", @stack;
}
[download]

I'm not really that familiar with Carp, I'll see what I can figure out to make it give a more complete backtrace.

Comment on Re^2: Request help for work around for bug in perl 5.10.1 (weak magic) Download Code

Replies are listed 'Best First'.
Re^3: Request help for work around for bug in perl 5.10.1 (backtrace) by tye (Sage) on Dec 12, 2012 at 22:34 UTC
I'm not really that familiar with Carp, I'll see what I can figure out to make it give a more complete backtrace. The problem wasn't that the backtrace was unacceptable. The problem was that you simply failed to log `@_` (really just `$_[0]`), which is the most important thing to log when logging inside of a __DIE__ handler. Luckily, the backtrace also happened to include a truncated version of that message. But the solution isn't to try to get the backtrace to not truncate. The solution is to log `$_[0]` directly (along with the backtrace). In this particular case, the interesting backtrace is the C backtrace, but that isn't as easy to get (you'd likely need a non-stripped build and gcc and...). - tye	[reply] [d/l] [select]

Replies are listed 'Best First'.

Re^3: Request help for work around for bug in perl 5.10.1 (backtrace)
by tye (Sage) on Dec 12, 2012 at 22:34 UTC

I'm not really that familiar with Carp, I'll see what I can figure out to make it give a more complete backtrace.

The problem wasn't that the backtrace was unacceptable. The problem was that you simply failed to log @_ (really just $_[0]), which is the most important thing to log when logging inside of a __DIE__ handler.

Luckily, the backtrace also happened to include a truncated version of that message. But the solution isn't to try to get the backtrace to not truncate. The solution is to log $_[0] directly (along with the backtrace).

In this particular case, the interesting backtrace is the C backtrace, but that isn't as easy to get (you'd likely need a non-stripped build and gcc and...).

- tye

[reply]
[d/l]
[select]