Referencing a PADTMP twice produces two copies

[p5pRT]

Migrated from rt.perl.org#78194 (status was 'resolved')

Searchable as RT78194$

[p5pRT]

From @cpansprout

$ perl -le' for( "$y" ){print \$_.\$_}'
SCALAR(0x8039f0)SCALAR(0x8038d0)

(The same address should be printed twice.)

It looks as though the TARG ought to be stolen by S_refto, and then replaced via pad_swipe or something similar.

---

Flags​:
  category=core
  severity=low

---

Site configuration information for perl 5.13.5​:

Configured by sprout at Mon Sep 27 23​:15​:22 PDT 2010.

Summary of my perl5 (revision 5 version 13 subversion 5) configuration​:
  Commit id​: 3eab78e
  Platform​:
  osname=darwin, osvers=10.4.0, archname=darwin-thread-multi-2level
  uname='darwin pint.local 10.4.0 darwin kernel version 10.4.0​: fri apr 23 18​:28​:53 pdt 2010; root​:xnu-1504.7.4~1release_i386 i386 '
  config_args='-de -Dusedevel -Duseithreads -DDEBUGGING'
  hint=recommended, useposix=true, d_sigaction=define
  useithreads=define, usemultiplicity=define
  useperlio=define, d_sfio=undef, uselargefiles=define, usesocks=undef
  use64bitint=undef, use64bitall=undef, uselongdouble=undef
  usemymalloc=n, bincompat5005=undef
  Compiler​:
  cc='cc', ccflags ='-fno-common -DPERL_DARWIN -no-cpp-precomp -DDEBUGGING -fno-strict-aliasing -pipe -fstack-protector -I/usr/local/include',
  optimize='-O3 -g',
  cppflags='-no-cpp-precomp -fno-common -DPERL_DARWIN -no-cpp-precomp -DDEBUGGING -fno-strict-aliasing -pipe -fstack-protector -I/usr/local/include'
  ccversion='', gccversion='4.2.1 (Apple Inc. build 5664)', gccosandvers=''
  intsize=4, longsize=4, ptrsize=4, doublesize=8, byteorder=1234
  d_longlong=define, longlongsize=8, d_longdbl=define, longdblsize=16
  ivtype='long', ivsize=4, nvtype='double', nvsize=8, Off_t='off_t', lseeksize=8
  alignbytes=8, prototype=define
  Linker and Libraries​:
  ld='env MACOSX_DEPLOYMENT_TARGET=10.3 cc', ldflags =' -fstack-protector -L/usr/local/lib'
  libpth=/usr/local/lib /usr/lib
  libs=-ldbm -ldl -lm -lutil -lc
  perllibs=-ldl -lm -lutil -lc
  libc=/usr/lib/libc.dylib, so=dylib, useshrplib=false, libperl=libperl.a
  gnulibc_version=''
  Dynamic Linking​:
  dlsrc=dl_dlopen.xs, dlext=bundle, d_dlsymun=undef, ccdlflags=' '
  cccdlflags=' ', lddlflags=' -bundle -undefined dynamic_lookup -L/usr/local/lib -fstack-protector'

Locally applied patches​:
 

---

@​INC for perl 5.13.5​:
  /usr/local/lib/perl5/site_perl/5.13.5/darwin-thread-multi-2level
  /usr/local/lib/perl5/site_perl/5.13.5
  /usr/local/lib/perl5/5.13.5/darwin-thread-multi-2level
  /usr/local/lib/perl5/5.13.5
  /usr/local/lib/perl5/site_perl
  .

---

Environment for perl 5.13.5​:
  DYLD_LIBRARY_PATH (unset)
  HOME=/Users/sprout
  LANG=en_US.UTF-8
  LANGUAGE (unset)
  LD_LIBRARY_PATH (unset)
  LOGDIR (unset)
  PATH=/usr/bin​:/bin​:/usr/sbin​:/sbin​:/usr/local/bin​:/usr/X11/bin​:/usr/local/bin
  PERL_BADLANG (unset)
  SHELL=/bin/bash

[p5pRT]

From @cpansprout

On Sun Oct 03 14​:08​:37 2010, sprout wrote​:

$ perl -le' for( "$y" ){print \$_.\$_}'
SCALAR(0x8039f0)SCALAR(0x8038d0)

(The same address should be printed twice.)

It looks as though the TARG ought to be stolen by S_refto, and then
replaced via pad_swipe or something similar.

It might be cleaner for not only S_refto, but also any op that aliases a
PADTMP to copy it. That would include for, map, grep, arguments to
subroutines, and the lhs of (...)x...

[p5pRT]

From [Unknown Contact. See original ticket]

On Sun Oct 03 14​:08​:37 2010, sprout wrote​:

$ perl -le' for( "$y" ){print \$_.\$_}'
SCALAR(0x8039f0)SCALAR(0x8038d0)

(The same address should be printed twice.)

It looks as though the TARG ought to be stolen by S_refto, and then
replaced via pad_swipe or something similar.

It might be cleaner for not only S_refto, but also any op that aliases a
PADTMP to copy it. That would include for, map, grep, arguments to
subroutines, and the lhs of (...)x...

[p5pRT]

@cpansprout - Status changed from 'new' to 'open'

[p5pRT]

From @cpansprout

On Sun Jun 05 16​:04​:53 2011, sprout wrote​:

On Sun Oct 03 14​:08​:37 2010, sprout wrote​:

$ perl -le' for( "$y" ){print \$_.\$_}'
SCALAR(0x8039f0)SCALAR(0x8038d0)

(The same address should be printed twice.)

It looks as though the TARG ought to be stolen by S_refto, and then
replaced via pad_swipe or something similar.

It might be cleaner for not only S_refto, but also any op that aliases a
PADTMP to copy it. That would include for, map, grep, arguments to
subroutines, and the lhs of (...)x...

Here is a difficult case​:

$ ./perl -Ilib -e 'use Devel​::Peek; "$_" =~ /(?{ Dump $_ })/;'
SV = PVMG(0x824b78) at 0x826ba0
  REFCNT = 1
  FLAGS = (PADTMP,SMG,POK,pPOK)
...

Fixing the operators I mentioned above is easy. Fixing patterns without
a speed hit (since regexps are used so often) is not so easy.

Are regular expressions with code blocks in them flagged somehow, so we
can avoid making a copy for every PADTMP =~ // ?

In case it’s not obvious, this is the bug I’m trying to fix​:

$ ./perl -Ilib -le '"$_" =~ /(?{ print \$_; print \$_ })/;'
SCALAR(0x803a10)
SCALAR(0x803b50)

Taking a reference to $_ twice should not create references to two
different variables. The implementation details are leaking through.

--

Father Chrysostomos

[p5pRT]

From [Unknown Contact. See original ticket]

On Sun Jun 05 16​:04​:53 2011, sprout wrote​:

On Sun Oct 03 14​:08​:37 2010, sprout wrote​:

$ perl -le' for( "$y" ){print \$_.\$_}'
SCALAR(0x8039f0)SCALAR(0x8038d0)

(The same address should be printed twice.)

It looks as though the TARG ought to be stolen by S_refto, and then
replaced via pad_swipe or something similar.

It might be cleaner for not only S_refto, but also any op that aliases a
PADTMP to copy it. That would include for, map, grep, arguments to
subroutines, and the lhs of (...)x...

Here is a difficult case​:

$ ./perl -Ilib -e 'use Devel​::Peek; "$_" =~ /(?{ Dump $_ })/;'
SV = PVMG(0x824b78) at 0x826ba0
  REFCNT = 1
  FLAGS = (PADTMP,SMG,POK,pPOK)
...

Fixing the operators I mentioned above is easy. Fixing patterns without
a speed hit (since regexps are used so often) is not so easy.

Are regular expressions with code blocks in them flagged somehow, so we
can avoid making a copy for every PADTMP =~ // ?

In case it’s not obvious, this is the bug I’m trying to fix​:

$ ./perl -Ilib -le '"$_" =~ /(?{ print \$_; print \$_ })/;'
SCALAR(0x803a10)
SCALAR(0x803b50)

Taking a reference to $_ twice should not create references to two
different variables. The implementation details are leaking through.

--

Father Chrysostomos

[p5pRT]

From @iabyn

On Sun, Jan 08, 2012 at 02​:21​:34PM -0800, Father Chrysostomos via RT wrote​:

On Sun Jun 05 16​:04​:53 2011, sprout wrote​:

On Sun Oct 03 14​:08​:37 2010, sprout wrote​:

$ perl -le' for( "$y" ){print \$_.\$_}'
SCALAR(0x8039f0)SCALAR(0x8038d0)

(The same address should be printed twice.)

It looks as though the TARG ought to be stolen by S_refto, and then
replaced via pad_swipe or something similar.

It might be cleaner for not only S_refto, but also any op that aliases a
PADTMP to copy it. That would include for, map, grep, arguments to
subroutines, and the lhs of (...)x...

Here is a difficult case​:

$ ./perl -Ilib -e 'use Devel​::Peek; "$_" =~ /(?{ Dump $_ })/;'
SV = PVMG(0x824b78) at 0x826ba0
REFCNT = 1
FLAGS = (PADTMP,SMG,POK,pPOK)
...

Fixing the operators I mentioned above is easy. Fixing patterns without
a speed hit (since regexps are used so often) is not so easy.

Are regular expressions with code blocks in them flagged somehow, so we
can avoid making a copy for every PADTMP =~ // ?

yes, RXf_EVAL_SEEN. It's used at the start of S_regtry to decide whether
to alias $_ (and PL_reg_eval_set is then set, to avoid doing it multiple
times).

In case it’s not obvious, this is the bug I’m trying to fix​:

$ ./perl -Ilib -le '"$_" =~ /(?{ print \$_; print \$_ })/;'
SCALAR(0x803a10)
SCALAR(0x803b50)

Taking a reference to $_ twice should not create references to two
different variables. The implementation details are leaking through.

Confused the hell out me! :-) I didn't realise until I looked at it just
now, that $_ is aliased within code blocks.

--
Gravity is just a theory; teach Intelligent Falling in our schools!
  http​://www.theonion.com/content/node/39512

[p5pRT]

From patcat88@snet.net

On Sun Jan 08 14​:21​:34 2012, sprout wrote​:
................................

Taking a reference to $_ twice should not create references to two
different variables. The implementation details are leaking through.

I'll add my PerlMonks discussion to this
http​://www.perlmonks.org/?node_id=941973 . PADTMPs should always be
copied (or sv body stolen ??), not referenced. srefgen/S_refto and newRV
do different things but they should be the same. PADTMPs are supposed to
be allocated and dealloced on block enterance/leaving, but under the
hood are effectively closures or C function level static. Its a mess.

[p5pRT]

From [Unknown Contact. See original ticket]

On Sun Jan 08 14​:21​:34 2012, sprout wrote​:
................................

Taking a reference to $_ twice should not create references to two
different variables. The implementation details are leaking through.

I'll add my PerlMonks discussion to this
http​://www.perlmonks.org/?node_id=941973 . PADTMPs should always be
copied (or sv body stolen ??), not referenced. srefgen/S_refto and newRV
do different things but they should be the same. PADTMPs are supposed to
be allocated and dealloced on block enterance/leaving, but under the
hood are effectively closures or C function level static. Its a mess.

[p5pRT]

From @cpansprout

This is related to #7505, which is one instance of this bug.

[p5pRT]

From [Unknown Contact. See original ticket]

This is related to #7505, which is one instance of this bug.

[p5pRT]

From @cpansprout

On Sat Jun 15 13​:49​:57 2013, sprout wrote​:

This is related to #7505, which is one instance of this bug.

Um, nope, it’s not. Ignore that.

--

Father Chrysostomos

[p5pRT]

From [Unknown Contact. See original ticket]

On Sat Jun 15 13​:49​:57 2013, sprout wrote​:

This is related to #7505, which is one instance of this bug.

Um, nope, it’s not. Ignore that.

--

Father Chrysostomos

[p5pRT]

From @cpansprout

Fixed in​:

a0ed822
b479c9f
8e079c2
706a6eb
da9e430
2b66f6d
82c2360

--

Father Chrysostomos

[p5pRT]

From [Unknown Contact. See original ticket]

Fixed in​:

a0ed822
b479c9f
8e079c2
706a6eb
da9e430
2b66f6d
82c2360

--

Father Chrysostomos