From @cpansprout

See https://rt.cpan.org/Ticket/Display.html?id=67239 and​:

commit 8165fae
Author​: Nicholas Clark <nick@​ccl4.org>
Date​: Sun Apr 18 13​:24​:20 2010 +0100

  Move PERL_ASYNC_CHECK() from POPBLOCK() to the kill case of Perl_apply().
 
  This ensures that (safe) signals sent to the same process are still dispatched
  within the same statement (as before), without overloading the semantics of
  block popping.

Flags​:
  category=core
  severity=high

Site configuration information for perl 5.14.0​:

Configured by sprout at Sun Apr 3 16​:29​:04 PDT 2011.

Summary of my perl5 (revision 5 version 14 subversion 0) configuration​:
  Snapshot of​: aa6faf7
  Platform​:
  osname=darwin, osvers=10.5.0, archname=darwin-thread-multi-2level
  uname='darwin pint.local 10.5.0 darwin kernel version 10.5.0​: fri nov 5 23​:20​:39 pdt 2010; root​:xnu-1504.9.17~1release_i386 i386 '
  config_args='-Dusedevel -de -Duseithreads -Doptimize=-g'
  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 -fno-strict-aliasing -pipe -fstack-protector -I/usr/local/include',
  optimize='-g',
  cppflags='-no-cpp-precomp -fno-common -DPERL_DARWIN -no-cpp-precomp -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​:
  RC0

@​INC for perl 5.14.0​:
  /usr/local/lib/perl5/site_perl/5.14.0/darwin-thread-multi-2level
  /usr/local/lib/perl5/site_perl/5.14.0
  /usr/local/lib/perl5/5.14.0/darwin-thread-multi-2level
  /usr/local/lib/perl5/5.14.0
  /usr/local/lib/perl5/site_perl
  .

Environment for perl 5.14.0​:
  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

From @cpansprout

Can this be made a blocker?

From @obra

On Sun 17.Apr'11 at 17​:18​:31 -0700, Father Chrysostomos wrote​:

Can this be made a blocker?

It now blocks. Thanks.

The RT System itself - Status changed from 'new' to 'open'

From @iabyn

On Mon, Apr 18, 2011 at 12​:11​:07PM +1000, Jesse Vincent wrote​:

On Sun 17.Apr'11 at 17​:18​:31 -0700, Father Chrysostomos wrote​:

Can this be made a blocker?

It now blocks. Thanks.

Now fixed​:

commit 011c381
Author​: David Mitchell <davem@​iabyn.com>
AuthorDate​: Tue Apr 19 14​:17​:12 2011 +0100
Commit​: David Mitchell <davem@​iabyn.com>
CommitDate​: Tue Apr 19 14​:32​:42 2011 +0100

  dispatch signals when leaving an eval
 
  Currently PERL_ASYNC_CHECK is only called during scope exit in pp_leavetry
  and pp_levaeeval. This means that if the signal handler calls die, the
  eval won't catch it.
 
  This broke Sys​::AlarmCall's test suite, which was doing the equivalent of
 
  $SIG{ALRM} = sub { die };
  eval {
  alarm(1);
  select(undef, undef, undef, 10);
  }
  # expect the die to get caught and $@​ set here.
 
  Because the select was the last statement in the block, PERL_ASYNC_CHECK
  wasn't called next until the leave_scope at the end of leavetry.
  See RT #88774.
 
  The simple fix is to add a PERL_ASYNC_CHECK at the top of
  leavetry and leaveeval.

--
"Do not dabble in paradox, Edward, it puts you in danger of fortuitous wit."
  -- Lady Croom, "Arcadia"

@iabyn - Status changed from 'open' to 'resolved'

From @nwc10

On Tue, Apr 19, 2011 at 02​:34​:09PM +0100, Dave Mitchell wrote​:

On Mon, Apr 18, 2011 at 12​:11​:07PM +1000, Jesse Vincent wrote​:

On Sun 17.Apr'11 at 17​:18​:31 -0700, Father Chrysostomos wrote​:

Can this be made a blocker?

It now blocks. Thanks.

Now fixed​:

Thanks. Your timing is perfect. I'd got the test case down to the same size as
yours, and had just started scratching my head about what was going on.
(Q​: "The signal clearly triggers inside the eval. So why on earth is the signal
handler getting run outside the eval?")

I'm glad that the fix is clean and simple, as I was concerned that it was
going to need something rather more "funky" (and hence risky)

(A​: Because if a PERL_ASYNC_CHECK() is missing from somewhere it should be,
the symptoms are equivalent to a signal being deferred due to a mask)

Nicholas Clark