Skip to content
Commits on Nov 23, 2015
  1. @wchristian
Commits on Jul 1, 2015
  1. v0.200013

            - document that Moo, rather than Moose, may be used if preferred
              (Brian Manning)
    committed Jul 1, 2015
  2. @spicyjack
  3. Throwable/Error.pm: Change POD, class is now built on Moo

    Brian Manning committed with Dec 17, 2014
Commits on Mar 13, 2015
  1. @oalders
Commits on Sep 20, 2014
  1. v0.200012

            - avoid memory leaks and infinite recursion when stringifying
              (thanks, Dagfinn Ilmari Mannsåker!)
    committed Sep 19, 2014
  2. @ilmari

    Disarm alarm after use

    ilmari committed with Apr 3, 2014
  3. @ilmari

    Avoid infinitely-recursive stringification

    Use new filter_frames_early option to Devel::StackTrace to ensure
    our own frames are filtered out before stringification happens.
    ilmari committed with Apr 1, 2014
  4. @ilmari

    Fix passing stack_trace_class to constructor

    Make the stack_trace attribute lazy and call it before BUILD.
    
    Also call ->as_string on the stack trace so that the raw stack trace
    gets filtered and the BUILD frame discarded, avoiding memory cycles on
    the exception object.
    ilmari committed with Apr 1, 2014
Commits on Apr 3, 2014
  1. v0.200011

            - eliminate prereq on Class::Load (thanks, Toby Inkster!)
            - eliminate preqre on MooX::Types::MooseLike::Base (Toby Inkster)
    committed Apr 2, 2014
  2. @tobyink

    drop dependency on MooX::Types::MooseLike::Base

    ...it was barely being used anyway
    
    Conflicts:
    	lib/StackTrace/Auto.pm
    tobyink committed with Mar 31, 2014
  3. correct diagnostic message

    committed Apr 2, 2014
  4. @tobyink

    Drop dependency on Class::Load.

    Moose has been moving away from it and towards Module::Runtime;
    Moo also uses Module::Runtime.
    
    For the stack trace class, rather than checking is_class_loaded,
    check $class->can("new"), which is more useful as we're going to
    call $class->new later.
    tobyink committed with Mar 31, 2014
Commits on Mar 9, 2014
  1. v0.200010

            - eliminate a memory leak in StackTrace::Auto (thanks, Graham Knop!)
    committed Mar 8, 2014
  2. prep for next release

    committed Mar 8, 2014
  3. @haarg

    fix memory leak

    haarg committed with Mar 6, 2014
  4. @haarg

    fix memory leak test

    haarg committed with Mar 6, 2014
Commits on Oct 22, 2013
  1. v0.200009

            - avoid boolifying objects that might be held in the out of band
              storage
            - update bugtracker and repo metadata
            - more changes coming soon
    committed Oct 21, 2013
  2. avoid boolifying potential objects

    committed Oct 21, 2013
Commits on Apr 30, 2013
  1. v0.200008

              avoid a Moose-using test in the suite of tests run by users (thanks,
              Dave Rolsky)
    committed Apr 30, 2013
  2. changelog

    committed Apr 30, 2013
  3. @autarch
  4. @autarch
Commits on Apr 9, 2013
  1. v0.200007

              allow previous_exception to be passed to the constructor
    committed Apr 9, 2013
  2. v0.200006

              lose old $@ in previous_exception less often under 5.17.10 (thanks,
              David Wheeler)
    committed Apr 9, 2013
  3. when doing a ->throw, pass $@ out of band

    Right now, the "previous_exception" attribute is unreliable, but you'd
    hardly know it unless you're using 5.17.10.
    
    The previous_exception attribute is supposed to store what $@ was when
    the Throwable object was created.  If previous_exception isn't supplied
    in the call to ->new, it defaults to whatever is in $@.  In almost every
    case, it's this default that gets used.  Nobody's likely to pass in the
    previous exception.
    
    If the previous_exception default runs early, it finds the old value of
    $@ intact and saves it.  If some other attribute's default runs first,
    that default sub might do an eval and clobber $@.  Then when the
    previous_exception default runs, it finds $@ empty.
    
    Attributes are initialized in an unreliable order.  If you look under
    the hood, you see it's hash traversal order over the hash that stores
    all the class's attributes.  In 5.17.10, hash traversal order got much
    less predictable.  Under 5.17.10, about half the time previous_exception
    was clobbered when it had been correct previously.
    
    This commit adds a lame hack to get a copy of $@ stored dynamically out
    of band during a call to SomethingThrowable->throw.  It helps, but it's
    not a panacea.  For one thing, ->new is still subject to the attribute
    ordering.
    
    We can't just put this in BUILDARGS.  For one thing, it's too likely to
    get composed with (say) MooseX::OneArgNew which will result in an
    unpredictable ordering of the "around" modifiers, presumably getting us
    right back to this problem.
    
    Maybe we can "around new," but for now this should be an improvement.
    committed Apr 9, 2013
  4. @theory

    Safer exception-throwing.

    theory committed Apr 9, 2013
  5. @theory

    Add t/strace_bug.t.

    On Perl 5.17.0, `previous_exception` is intermittently undefined. Note that
    the test uses its own stack trace class that does nothing, not
    Devel::StackTrace.
    theory committed Apr 9, 2013
Commits on Apr 5, 2013
  1. v0.200005

              Plugged memory leak (thaljef)
    committed Apr 5, 2013
Commits on Mar 25, 2013
  1. Update change log.

    Jeffrey Ryan Thalhammer committed Mar 25, 2013
Something went wrong with that request. Please try again.