[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index][Thread Index][Top&Search][Original]
[ID 20000111.001] carp/croak may not show caller's module/line
This is a bug report for perl from Wolfgang.Laun@alcatel.at,
generated with the help of perlbug 1.26 running under perl 5.00503.
-----------------------------------------------------------------
[Please enter your report here]
Perl bug report concerning the library module Carp
This concerns Carp::Heavy.pm (or Carp.pm) in all perl 5.x versions,
including the latest developer version.
*** Bug: carp and croak may fail to show "caller"
Carp's carp and croak used in some module M should report errors from
a caller's point of view. As the comment in Heavy.pm says, a "caller"
is some module not among M's ancestors or descendants.
Consider this set of modules (on file Animal.pm):
#========================================================
package Animal;
use UNIVERSAL qw( can );
use Carp;
sub new {
my $class = shift;
bless( { @_ }, $class );
}
sub dump {
my $self = shift();
$self->name() if $self->can(name);
}
sub teeth {
my $self = shift();
carp( "Has no teeth" ) unless $self->{-teeth};
}
#========================================================
package Mammal;
@ISA = ( 'Animal' );
sub new {
my $class = shift;
bless( $class->SUPER::new( @_ ), $class );
}
#========================================================
package Rodent;
use Carp;
@ISA = ( 'Mammal' );
sub new {
my $class = shift();
bless( $class->SUPER::new( @_ ), $class );
}
sub name {
my $self = shift();
carp( "Name called - no name" ) unless $self->{-name};
}
sub gnaw {
my $self = shift();
$self->teeth();
}
#========================================================
Now, let's run this program:
use Animal;
my $r = Rodent->new( -name => '', -teeth => 0 );
$r->dump();
$r->gnaw();
Output:
Name called - no name at Animal.pm line 13
Has no teeth at Animal.pm line 47
carp reports lines within Animal.pm! Both the call of Rodent::name
from Animal::dump and the call of Animal::teeth from Rodent::gnaw are
not detected to be within M's own domain.
The algorithm in shortmess_heavy is faulty as it fails to look beyond
a module's immediate ancestors and inheritors. The enclosed version
of Carp::Heavy.pm does not show this defect.
While fixing this, the $CarpLevel feature required some consideration.
This has led to the question:
*** Is $CarpLevel really useful?
While not in the official documentation, the comments in Heavy.pm would
give the impression that increasing $CarpLevel would enable a wrapper
module to shift the reporting of carp/croak/confess/cluck to the "right"
place.
I doubt that this can be achieved in this simple manner. First, any
wrapper wanting to appear transparent to carp & co. would have to
enclose all calls to the wrapped module's subroutines in $CarpLevel++
and $CarpLevel-- statements. (Clearly this increment cannot be done
statically as it would affect all modules around.) Second, what would
happen if a wrapper increments $CarpLevel dynamically, calls the wrapped
module which does a callback to a user subroutine and this calls the
wrapped module again, which then calls carp?
To make this feature robust, we'd need a call telling Carp to treat a
wrapper module W as "one of the family" with respect to some wrapped
module M, e.g.
Carp::Adopt( M, W );
If some M subroutine carps, and the chain of callers passes into
W's family tree, the algorithm continues.
This probably should also influence the longmess_heavy algorithm.
(As this goes way beyond a simple bugfix, I've refrained from touching
this.)
==== Carp::Heavy.pm ======================================================
package Carp;
=head1 NAME
Carp::Heavy - Carp guts
=head1 SYNOPIS
(internal use only)
=head1 DESCRIPTION
No user-serviceable parts inside.
=cut
# This package is heavily used. Be small. Be fast. Be good.
# Comments added by Andy Wardley <abw@kfs.org> 09-Apr-98, based on an
# _almost_ complete understanding of the package. Corrections and
# comments are welcome.
# longmess() crawls all the way up the stack reporting on all the function
# calls made. The error string, $error, is originally constructed from the
# arguments passed into longmess() via confess(), cluck() or shortmess().
# This gets appended with the stack trace messages which are generated for
# each function call on the stack.
sub longmess_heavy {
return @_ if ref $_[0];
my $error = join '', @_;
my $mess = "";
my $i = 1 + $CarpLevel;
my ($pack,$file,$line,$sub,$hargs,$eval,$require);
my (@a);
#
# crawl up the stack....
#
while (do { { package DB; @a = caller($i++) } } ) {
# get copies of the variables returned from caller()
($pack,$file,$line,$sub,$hargs,undef,$eval,$require) = @a;
#
# if the $error error string is newline terminated then it
# is copied into $mess. Otherwise, $mess gets set (at the end of
# the 'else {' section below) to one of two things. The first time
# through, it is set to the "$error at $file line $line" message.
# $error is then set to 'called' which triggers subsequent loop
# iterations to append $sub to $mess before appending the "$error
# at $file line $line" which now actually reads "called at $file line
# $line". Thus, the stack trace message is constructed:
#
# first time: $mess = $error at $file line $line
# subsequent times: $mess .= $sub $error at $file line $line
# ^^^^^^
# "called"
if ($error =~ m/\n$/) {
$mess .= $error;
} else {
# Build a string, $sub, which names the sub-routine called.
# This may also be "require ...", "eval '...' or "eval {...}"
if (defined $eval) {
if ($require) {
$sub = "require $eval";
} else {
$eval =~ s/([\\\'])/\\$1/g;
if ($MaxEvalLen && length($eval) > $MaxEvalLen) {
substr($eval,$MaxEvalLen) = '...';
}
$sub = "eval '$eval'";
}
} elsif ($sub eq '(eval)') {
$sub = 'eval {...}';
}
# if there are any arguments in the sub-routine call, format
# them according to the format variables defined earlier in
# this file and join them onto the $sub sub-routine string
if ($hargs) {
# we may trash some of the args so we take a copy
@a = @DB::args; # must get local copy of args
# don't print any more than $MaxArgNums
if ($MaxArgNums and @a > $MaxArgNums) {
# cap the length of $#a and set the last element to '...'
$#a = $MaxArgNums;
$a[$#a] = "...";
}
for (@a) {
# set args to the string "undef" if undefined
$_ = "undef", next unless defined $_;
if (ref $_) {
# force reference to string representation
$_ .= '';
s/'/\\'/g;
}
else {
s/'/\\'/g;
# terminate the string early with '...' if too long
substr($_,$MaxArgLen) = '...'
if $MaxArgLen and $MaxArgLen < length;
}
# 'quote' arg unless it looks like a number
$_ = "'$_'" unless /^-?[\d.]+$/;
# print high-end chars as 'M-<char>'
s/([\200-\377])/sprintf("M-%c",ord($1)&0177)/eg;
# print remaining control chars as ^<char>
s/([\0-\37\177])/sprintf("^%c",ord($1)^64)/eg;
}
# append ('all', 'the', 'arguments') to the $sub string
$sub .= '(' . join(', ', @a) . ')';
}
# here's where the error message, $mess, gets constructed
$mess .= "\t$sub " if $error eq "called";
$mess .= "$error at $file line $line";
if (defined &Thread::tid) {
my $tid = Thread->self->tid;
$mess .= " thread $tid" if $tid;
}
$mess .= "\n";
}
# we don't need to print the actual error message again so we can
# change this to "called" so that the string "$error at $file line
# $line" makes sense as "called at $file line $line".
$error = "called";
}
# this kludge circumvents die's incorrect handling of NUL
my $msg = \($mess || $error);
$$msg =~ tr/\0//d;
$$msg;
}
# ancestors() returns the complete set of ancestors of a module
sub ancestors($$){
my( $pack, $href ) = @_;
my $risa = \@{"${pack}::ISA"};
if( @$risa ){
my %tree = ();
@tree{@$risa} = ();
foreach my $mod ( @$risa ){
# visit ancestors - if not already in the gallery
if( ! defined( $$href{$mod} ) ){
my @ancs = ancestors( $mod, $href );
@tree{@ancs} = ();
}
}
return ( keys( %tree ) );
} else {
return ();
}
}
# shortmess() is called by carp() and croak() to skip all the way up to
# the top-level caller's package and report the error from there. confess()
# and cluck() generate a full stack trace so they call longmess() to
# generate that. In verbose mode shortmess() calls longmess() so
# you always get a stack trace
sub shortmess_heavy { # Short-circuit &longmess if called via multiple packages
goto &longmess_heavy if $Verbose;
return @_ if ref $_[0];
my $error = join '', @_;
my ($prevpack) = caller(1);
my $extra = $CarpLevel;
my @Clans = ( $prevpack );
my $i = 2;
my ($pack,$file,$line);
# when reporting an error, we want to report it from the context of the
# calling package. So what is the calling package? Within a module,
# there may be many calls between methods and perhaps between sub-classes
# and super-classes, but the user isn't interested in what happens
# inside the package. We start by building a hash array which keeps
# track of all the packages to which the calling package belongs. We
# do this by examining its @ISA variable. Any call from a base class
# method (one of our caller's @ISA packages) can be ignored
my %isa;
# merge all the caller's @ISA packages and ancestors into %isa.
my @pars = ancestors( $prevpack, \%isa );
@isa{@pars} = () if @pars;
$isa{$prevpack} = 1;
# now we crawl up the calling stack and look at all the packages in
# there. For each package, we look to see if it has an @ISA and then
# we see if our caller features in that list. That would imply that
# our caller is a derived class of that package and its calls can also
# be ignored
CALLER:
while (($pack,$file,$line) = caller($i++)) {
# Chances are, the caller's caller (or its caller...) is already
# in the gallery - if so, ignore this caller.
next if exists( $isa{$pack} );
# no: collect this module's ancestors.
my @i = ancestors( $pack, \%isa );
my %i;
if( @i ){
@i{@i} = ();
# check whether our representative of one of the clans is
# in this family tree.
foreach my $cl (@Clans){
if( exists( $i{$cl} ) ){
# yes: merge all of the family tree into %isa
@isa{@i,$pack} = ();
# and here's where we do some more ignoring...
# if the package in question is one of our caller's
# base or derived packages then we can ignore it (skip it)
# and go onto the next.
next CALLER if exists( $isa{$pack} );
last;
}
}
}
# Hey! We've found a package that isn't one of our caller's
# clan....but wait, $extra refers to the number of 'extra' levels
# we should skip up. If $extra > 0 then this is a false alarm.
# We must merge the package into the %isa hash (so we can ignore it
# if it pops up again), decrement $extra, and continue.
if ($extra-- > 0) {
push( @Clans, $pack );
@isa{@i,$pack} = ();
}
else {
# OK! We've got a candidate package. Time to construct the
# relevant error message and return it. die() doesn't like
# to be given NUL characters (which $msg may contain) so we
# remove them first.
my $msg;
$msg = "$error at $file line $line";
if (defined &Thread::tid) {
my $tid = Thread->self->tid;
$mess .= " thread $tid" if $tid;
}
$msg .= "\n";
$msg =~ tr/\0//d;
return $msg;
}
}
# uh-oh! It looks like we crawled all the way up the stack and
# never found a candidate package. Oh well, let's call longmess
# to generate a full stack trace. We use the magical form of 'goto'
# so that this shortmess() function doesn't appear on the stack
# to further confuse longmess() about it's calling package.
goto &longmess_heavy;
}
1;
[Please do not change anything below this line]
-----------------------------------------------------------------
---
Site configuration information for perl 5.00503:
Configured by root at Mon Aug 30 23:08:56 EDT 1999.
Summary of my perl5 (5.0 patchlevel 5 subversion 3) configuration:
Platform:
osname=linux, osvers=2.2.5-22smp, archname=i386-linux
uname='linux porky.devel.redhat.com 2.2.5-22smp #1 smp wed jun 2 09:11:51 edt 1999 i686 unknown '
hint=recommended, useposix=true, d_sigaction=define
usethreads=undef useperlio=undef d_sfio=undef
Compiler:
cc='cc', optimize='-O2', gccversion=egcs-2.91.66 19990314/Linux (egcs-1.1.2 release)
cppflags='-Dbool=char -DHAS_BOOL -I/usr/local/include'
ccflags ='-Dbool=char -DHAS_BOOL -I/usr/local/include'
stdchar='char', d_stdstdio=undef, usevfork=false
intsize=4, longsize=4, ptrsize=4, doublesize=8
d_longlong=define, longlongsize=8, d_longdbl=define, longdblsize=12
alignbytes=4, usemymalloc=n, prototype=define
Linker and Libraries:
ld='cc', ldflags =' -L/usr/local/lib'
libpth=/usr/local/lib /lib /usr/lib
libs=-lnsl -ldl -lm -lc -lposix -lcrypt
libc=, so=so, useshrplib=false, libperl=libperl.a
Dynamic Linking:
dlsrc=dl_dlopen.xs, dlext=so, d_dlsymun=undef, ccdlflags='-rdynamic'
cccdlflags='-fpic', lddlflags='-shared -L/usr/local/lib'
Locally applied patches:
---
@INC for perl 5.00503:
/usr/lib/perl5/5.00503/i386-linux
/usr/lib/perl5/5.00503
/usr/lib/perl5/site_perl/5.005/i386-linux
/usr/lib/perl5/site_perl/5.005
.
---
Environment for perl 5.00503:
HOME=/home/laune
LANG=en
LANGUAGE (unset)
LC_ALL=en_US
LD_LIBRARY_PATH (unset)
LOGDIR (unset)
PATH=/usr/bin:/bin:/usr/X11R6/bin:/usr/local/bin:/opt/bin:/usr/X11R6/bin:/home/laune/bin:/usr/X11R6/bin:/home/laune/bin
PERL_BADLANG (unset)
SHELL=/bin/bash
[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index][Thread Index][Top&Search][Original]