Perl
duplicate labels in the same block should be errors; duplicate labels in the branches of a conditional are not independent
Description
Two issues, one documentation, one behavior.
Documentation: perlsyn doesn't document the scope of labels and what happens if a label is redefined.
Behavior
Here's what happens if the loop labels are in the same scope:
#! /usr/bin/env perl
use v5.10;
my $x = 0;
my $LOOP;
LOOP: {
say defined $LOOP
? "goto $LOOP, got LOOP 1"
: 'Passing through LOOP 1';
}
$LOOP = 'LOOP 1';
goto LOOP if $x++ == 0;
undef $LOOP;
LOOP: {
say defined $LOOP
? "goto $LOOP, got LOOP 2"
: 'Passing through LOOP 2';
}
$LOOP = 'LOOP 2';
goto LOOP if $x++ == 2;
$ perl docbug.pl
Passing through LOOP 1
goto LOOP 1, got LOOP 1
Passing through LOOP 2
goto LOOP 2, got LOOP 1
Passing through LOOP 2
Looks like the repeated label is ignored. That's unexpected. It should be an error if one shouldn't repeat labels.
What if they are in separate blocks?
#! /usr/bin/env perl
use v5.10;
my $x = 0;
my $LOOP;
{
LOOP: {
say defined $LOOP
? "goto $LOOP, got LOOP 1"
: 'Passing through LOOP 1';
}
$LOOP = 'LOOP 1';
goto LOOP if $x++ == 0;
}
undef $LOOP;
{
LOOP: {
say defined $LOOP
? "goto $LOOP, got LOOP 2"
: 'Passing through LOOP 2';
}
$LOOP = 'LOOP 2';
goto LOOP if $x++ == 2;
}
$ perl docbug.pl
Passing through LOOP 1
goto LOOP 1, got LOOP 1
Passing through LOOP 2
goto LOOP 2, got LOOP 2
Looks like they are treated as separate labels, which is what I'd expect.
What if they are declared in blocks in a conditional?
#! /usr/bin/env perl
use v5.10;
for my $cond ( 0 .. 1 ) {
my $x = 0;
my $LOOP;
if ( 1 - $cond ) {
LOOP: {
say defined $LOOP
? "goto $LOOP, got LOOP 1"
: 'Passing through LOOP 1';
$LOOP = 'LOOP 1';
goto LOOP if $x++ == 0;
}
}
elsif ( $cond ) {
LOOP: {
say defined $LOOP
? "goto $LOOP, got LOOP 2"
: 'Passing through LOOP 2';
$LOOP = 'LOOP 2';
goto LOOP if $x++ == 0;
}
}
}
$ perl docbugif.pl
Passing through LOOP 1
goto LOOP 1, got LOOP 1
Passing through LOOP 2
goto LOOP 2, got LOOP 1
Oops; the second one is ignored, even though it's in a new block. That is definitely not what I expected. I expected this situation to mirror the second example.
Note that if the first branch of the conditional is optimized out (see below), the first label isn't defined, and the second branch sees its own label. This means the optimization is changing the behavior of code outside of the optimized out part, which shouldn't happen. I think the above behavior is a bug.
#! /usr/bin/env perl
use v5.10;
use strict;
use warnings;
for my $cond ( 0 .. 1 ) {
my $x = 0;
my $LOOP;
if ( 0 ) {
LOOP: {
say defined $LOOP
? "goto $LOOP, got LOOP 1"
: 'Passing through LOOP 1';
$LOOP = 'LOOP 1';
goto LOOP if $x++ == 0;
}
}
elsif ( $cond ) {
LOOP: {
say defined $LOOP
? "goto $LOOP, got LOOP 2"
: 'Passing through LOOP 2';
$LOOP = 'LOOP 2';
goto LOOP if $x++ == 0;
}
}
}
$ perl docbugif.pl
Passing through LOOP 2
goto LOOP 2, got LOOP 2
Expected behavior
- I expect that a repeated label in the same scope would be an error.
- I expect that labels in different blocks are independent.
- I expect that in a conditional where the label is defined in two separate blocks, they should be treated as separate labels; the first example violates that, and further creates a situation where optimizing out the first branch of the conditional affects code outside of it.
Perl configuration
Summary of my perl5 (revision 5 version 38 subversion 2) configuration:
Platform:
osname=linux
osvers=6.1.0-16-amd64
archname=x86_64-linux
uname='linux leafhopper 6.1.0-16-amd64 #1 smp preempt_dynamic debian 6.1.67-1 (2023-12-12) x86_64 gnulinux '
config_args='-Dprefix=/home/dj/.plenv/versions/5.38.2 -de -Dversiononly -A'eval:scriptdir=/home/dj/.plenv/versions/5.38.2/bin''
hint=recommended
useposix=true
d_sigaction=define
useithreads=undef
usemultiplicity=undef
use64bitint=define
use64bitall=define
uselongdouble=undef
usemymalloc=n
default_inc_excludes_dot=define
Compiler:
cc='cc'
ccflags ='-fwrapv -fno-strict-aliasing -pipe -fstack-protector-strong -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -D_FORTIFY_SOURCE=2'
optimize='-O2'
cppflags='-fwrapv -fno-strict-aliasing -pipe -fstack-protector-strong -I/usr/local/include'
ccversion=''
gccversion='12.2.0'
gccosandvers=''
intsize=4
longsize=8
ptrsize=8
doublesize=8
byteorder=12345678
doublekind=3
d_longlong=define
longlongsize=8
d_longdbl=define
longdblsize=16
longdblkind=3
ivtype='long'
ivsize=8
nvtype='double'
nvsize=8
Off_t='off_t'
lseeksize=8
alignbytes=8
prototype=define
Linker and Libraries:
ld='cc'
ldflags =' -fstack-protector-strong -L/usr/local/lib'
libpth=/usr/local/lib /usr/lib/x86_64-linux-gnu /usr/lib /usr/lib64
libs=-lpthread -ldl -lm -lcrypt -lutil -lc
perllibs=-lpthread -ldl -lm -lcrypt -lutil -lc
libc=/lib/x86_64-linux-gnu/libc.so.6
so=so
useshrplib=false
libperl=libperl.a
gnulibc_version='2.36'
Dynamic Linking:
dlsrc=dl_dlopen.xs
dlext=so
d_dlsymun=undef
ccdlflags='-Wl,-E'
cccdlflags='-fPIC'
lddlflags='-shared -O2 -L/usr/local/lib -fstack-protector-strong'
Characteristics of this binary (from libperl):
Compile-time options:
HAS_LONG_DOUBLE
HAS_STRTOLD
HAS_TIMES
PERLIO_LAYERS
PERL_COPY_ON_WRITE
PERL_DONT_CREATE_GVSV
PERL_HASH_FUNC_SIPHASH13
PERL_HASH_USE_SBOX32
PERL_MALLOC_WRAP
PERL_OP_PARENT
PERL_PRESERVE_IVUV
PERL_USE_SAFE_PUTENV
USE_64_BIT_ALL
USE_64_BIT_INT
USE_LARGE_FILES
USE_LOCALE
USE_LOCALE_COLLATE
USE_LOCALE_CTYPE
USE_LOCALE_NUMERIC
USE_LOCALE_TIME
USE_PERLIO
USE_PERL_ATOF
Built under linux
Compiled at Jan 18 2024 17:41:48
%ENV:
PERL_CPANM_OPT="--cascade-search --mirror-only --mirror /home/dj/Work/darkpan --mirror http://www.cpan.org"
PERL_CPAN_MIRROR_TINY_BASE="/home/dj/Work/darkpan"
PERL_DARKPAN="/home/dj/Work/darkpan"
@INC:
/home/dj/.plenv/versions/5.38.2/lib/perl5/site_perl/5.38.2/x86_64-linux
/home/dj/.plenv/versions/5.38.2/lib/perl5/site_perl/5.38.2
/home/dj/.plenv/versions/5.38.2/lib/perl5/5.38.2/x86_64-linux
/home/dj/.plenv/versions/5.38.2/lib/perl5/5.38.2
Duplicate labels in the same scope are commonly used in tests:
SKIP:
{
skip "I can't test this", 1 if $cant_test_this;
...
}
SKIP:
{
skip "I can't test that", 1 if $cant_test_that;
...
}
so we can't just warn on duplicate labels in the same scope.
This means the optimization is changing the behavior of code outside of the optimized out part, which shouldn't happen.
This is documented in perlsyn and perlfunc:
... It also can't be used to go into a construct that is optimized away. ...
Of course, there's no requirement that the label is defined in the function containing the goto:
tony@venus:.../git/perl6$ cat ../22088b.pl
use v5.38;
sub f {
say "f";
goto X;
}
f();
goto Y;
X: { say "X1"; }
goto Y;
X: { say "X2"; }
Y:
{
f();
X: say "X3";
}
tony@venus:.../git/perl6$ ./perl -Ilib ../22088b.pl
f
X1
f
X3
I do think how this is resolved could be better documented.
Duplicate labels in the same scope are commonly used in tests:
SKIP: { skip "I can't test this", 1 if $cant_test_this; ... } SKIP: { skip "I can't test that", 1 if $cant_test_that; ... }so we can't just warn on duplicate labels in the same scope.
Darn. Is there a way to differentiate between goto and the loop control statements that must be used within the block they address (skip uses last under the hood)? It'd have to be a run-time error, as goto can take an expression; maybe a flag on a label indicating it's multiply defined so that when goto uses it it knows it's degenerate.
This is documented in perlsyn and perlfunc:
... It also can't be used to go into a construct that is optimized away. ...
Well, so much for my reading comprehension. Thanks for pointing that bit out, somehow I glossed over that. Since there's no official "guide to what gets optimized away" perhaps that should be written in bold and prefaced with "Here be dragons".
One documentation tweak might be to suggest use of redo rather than goto when jumping to the beginning of a labeled block from within it. I've modified my code to do that, so the multiple definition issue is resolved.