Make the fat comma test more pathological.

[p5sagit/Devel-Declare.git] / Changes

Changes for Devel-Declare

0.005010 - 11 Aug 2009
  - Don't invoke the linestr callback if we found a keyword and the bufptr
    still contains an arrow at its beginning. This causes the linestr callback
    not to be fired on things like "->method" (if method is a declarator).
    While this might be useful, it mostly caused problems so far. It might be
    added again later as a separate feature.

0.005009 - 10 Aug 2009
  - Stop mixing declarations with code. (Closes: RT#48548) (Cosimo Streppone)
  - Move strip_attrs, which is a purely parsing method, from
    MethodInstaller::Simple to Context::Simple (nperez).

0.005008 - 27 Jul 2009
  - Depend on the latest ExtUtils::Depends for win32+gcc support.
  - Conditionally expand linestrings under perl debugger.

0.005007 - 13 Jul 2009
  - Fix line numbers not being updated from skipspace calls

0.005006 - 16 Jun 2009
  - Fix compilation on 5.8 perls.

0.005005 - 05 Jun 2009
  - Improve compatibility with MAD-enabled perls
    (Reini Urban, Closes RT#45779).

0.005004 - 03 Jun 2009
  - Don't redefine MEM_WRAP_CHECK_ if it's already defined, getting rid of
    compilation errors on some perls (Maik Fischer).

0.005003 - 24 May 2009
  - Failing tests for line number issues (Ash Berlin).
  - Add strip_names_and_args (Cory Watson).
  - Various pod fixes (Yanick Champoux, Florian Ragwitz).
  - Add copyright statements.

0.005002
  - Don't invoke the linestr callback if the parser was expecting an operator.
    This makes calling a method with the name of a declarator work.

0.005001
  - Implement skip_declarator in terms of scan_word in Context::Simple.
    This avoids relying on PL_tokenbuf to skip past the declarator, as
    PL_tokenbuf might be overwritten by compiling other code between calling
    the linestr callback and skip_declarator.

0.005000
  - Port the documentation from %^H and Scope::Guard to B::Hooks::EndOfScope.
  - Fix extracting multi-line strings using scan_str.
  - Remove the nasty workaround for the above from Context::Simple.

0.004000
  - Much improved documentation (osfameron).
  - Make sure the linestr callback isn't invoked when a declarator is used as
    part of some quoting construct like qq//, m//, s///, qr//, tr///, qx//, ``,
    etc (Florian Ragwitz).
    - Tests for this (osfameron, Florian Ragwitz).

0.003005
  - Depend on a recent B::Hooks::EndOfScope to make semicolon injection more
    robust (Florian Ragwitz).
  - Add a couple of TODO tests for better error reporting (Marcus Ramberg).
  - Context::Simple::inject_if_block now returns true if start of block was
    found (Ash Berlin).

0.003004
  - Make magic work within string evals on 5.10 if the hints hash is used.
  - Bind S_scan_ident to perl. It allows scanning for simple identifiers.
  - Add strip_ident to Context::Simple.

0.003003
  - Devel::Declare::MethodInstaller::Simple now has code_for() which the
    subclass can override to monkey with the magic shadowed subroutine.
    This is handy if you want to employ Devel::BeginLift.

0.003002
  - Depend on Sub::Name and B::Hooks::EndOfScope. MethodInstaller::Simple and
    Context::Simple use them.

0.003001
  - Don't use :lvalue in Context::Simple and MethodInstaller::Simple to keep
    the debugger happy.

0.003000
  - Add Devel::Declare::Context::Simple and
    Devel::Declare::MethodInstaller::Simple as an experimental way to make
    creating block- and sub-like keywords easier (Rhesa Rozendaal).

0.002999_01
  - Use B::Hooks::OP::Check to register PL_check callbacks.
  - Use B::Hooks::EndOfScope instead of %^H and Scope::Guard.
  - Don't segfault if HvNAME(PL_curstash) == NULL.
  - Don't segfault on 5.9.5+, where PL_parser is a symbol, not a define.
  - Don't delete the previous symbol table entry when shadowing subs. This
    makes us work within the debugger.
  - Don't mix declarations and code.

0.002002
  - switch done_declare call from call_argv to call_pv.
  - Make get_linestr{,_offset} return sensible values when called while the
    parser isn't running.
  - Remove several compile time warnings.
  - Allow enabling of debug mode using $ENV{DD_DEBUG}.

0.002001
  - clean up checks for whether we're lexing and whether lex_stuff exists
    to handle the PL_parser factor-out in 5.10
  - check if reallocation of PL_linestr is necessary before doing it. this way
    we can bail out properly instead of corrupting memory in some cases
  - don't call strlen twice on the same sting
  - try to be more portable
    - stop using Nullsv
    - don't use Perl_* functions directly.
    - don't define PERL_CORE
    - use NEWSV from handy.h instead of defining our own
    - don't define PERL_NO_GET_CONTEXT
    - don't support preprocessors (perl -P)

0.002000
  - rewrite guts into perl, add tests for new declaration style

0.001011
  - add support for 'method main' and other package names

0.001010
  - fix traits code, again, so it compiles on 5.10.

0.001009
  - only mangle evals if o->op_ppaddr is actually PL_ppaddr[OP_ENTERVAL]
  - don't set OPf_SPECIAL on entereval ops, mistaken cargo cult from autobox
  - fix traits code to work on older 5.8.x perls

0.001008
  - turns out 0.1.7 in Makefile.PL results in that version going into
    the .xs file as well as the dist name. Then everything breaks.

0.001007
  - bail on SvGROW during declare process since that can't work
  - use a source filter on files and an op wrap on evals to pre-grow
    PL_linestr so we don't get to the point of needing to bail

0.001006
  - nasty goto &$func hack to avoid :lvalue+list context weirdness
  - correct SvGROW invocation

0.001005
  - stop using & prototypes at all

0.001004
  - correct idiotic typo if ifndef

0.001003
  - internalise definitions for toke.c chunks to save hassle
  - make NEWSV macro defined via ifndef rather than version check in case
    of 5.8.8-maint (and 5.8.9)

0.001002
  - compilation fixes for 5.9.5 and -DDEBUGGING

0.001001
  - compilation fixes for 5.8.1+
  - set Makefile.PL to bomb out for <5.8.1

0.001000
  - Initial release to CPAN