These were reverted in gh-26530 (commit 17c4edc) due to refleaks.
* 2c1e258 - Compute deref offsets in compiler (gh-25152)
* b2bf2bc - Add new internal code objects fields: co_fastlocalnames and co_fastlocalkinds. (gh-26388)
This change fixes the refleaks.
https://bugs.python.org/issue43693
* Revert "bpo-43693: Compute deref offsets in compiler (gh-25152)"
This reverts commit b2bf2bc1ec.
* Revert "bpo-43693: Add new internal code objects fields: co_fastlocalnames and co_fastlocalkinds. (gh-26388)"
This reverts commit 2c1e2583fd.
These two commits are breaking the refleak buildbots.
Merges locals and cells into a single array.
Saves a pointer in the interpreter and means that we don't need the LOAD_CLOSURE opcode any more
https://bugs.python.org/issue43693
A number of places in the code base (notably ceval.c and frameobject.c) rely on mapping variable names to indices in the frame "locals plus" array (AKA fast locals), and thus opargs. Currently the compiler indirectly encodes that information on the code object as the tuples co_varnames, co_cellvars, and co_freevars. At runtime the dependent code must calculate the proper mapping from those, which isn't ideal and impacts performance-sensitive sections. This is something we can easily address in the compiler instead.
This change addresses the situation by replacing internal use of co_varnames, etc. with a single combined tuple of names in locals-plus order, along with a minimal array mapping each to its kind (local vs. cell vs. free). These two new PyCodeObject fields, co_fastlocalnames and co_fastllocalkinds, are not exposed to Python code for now, but co_varnames, etc. are still available with the same values as before (though computed lazily).
Aside from the (mild) performance impact, there are a number of other benefits:
* there's now a clear, direct relationship between locals-plus and variables
* code that relies on the locals-plus-to-name mapping is simpler
* marshaled code objects are smaller and serialize/de-serialize faster
Also note that we can take this approach further by expanding the possible values in co_fastlocalkinds to include specific argument types (e.g. positional-only, kwargs). Doing so would allow further speed-ups in _PyEval_MakeFrameVector(), which is where args get unpacked into the locals-plus array. It would also allow us to shrink marshaled code objects even further.
https://bugs.python.org/issue43693
"Zero cost" exception handling.
* Uses a lookup table to determine how to handle exceptions.
* Removes SETUP_FINALLY and POP_TOP block instructions, eliminating (most of) the runtime overhead of try statements.
* Reduces the size of the frame object by about 60%.
* Use instruction offset, rather than bytecode offset. Streamlines interpreter dispatch a bit, and removes most EXTENDED_ARGs for jumps.
* Change some uses of PyCode_Addr2Line to PyFrame_GetLineNumber
Document that lnotab can contain invalid bytecode offsets (because of
terrible reasons that are difficult to fix). Make dis.findlinestarts()
ignore invalid offsets in lnotab. All other uses of lnotab in CPython
(various reimplementations of addr2line or line2addr in Python, C and gdb)
already ignore this, because they take an address to look for, instead.
Add tests for the result of dis.findlinestarts() on wacky constructs in
test_peepholer.py, because it's the easiest place to add them.
This commit contains the implementation of PEP570: Python positional-only parameters.
* Update Grammar/Grammar with new typedarglist and varargslist
* Regenerate grammar files
* Update and regenerate AST related files
* Update code object
* Update marshal.c
* Update compiler and symtable
* Regenerate importlib files
* Update callable objects
* Implement positional-only args logic in ceval.c
* Regenerate frozen data
* Update standard library to account for positional-only args
* Add test file for positional-only args
* Update other test files to account for positional-only args
* Add News entry
* Update inspect module and related tests
Coroutines and async generators use a distinct attribute name for their
code objects, so this updates the `dis` module to correctly disassemble
objects with those attributes.
Due to the increase in the test module length, it also fixes some latent
defects in the tests related to how the displayed source line numbers
are extracted.
https://bugs.python.org/issue31230 is a follow-up issue suggesting we
may want to solve this a different way, by instead giving all these object
types a common `__code__` attribute, avoiding the need for special
casing in the `dis` module.
* BUILD_TUPLE_UNPACK and BUILD_MAP_UNPACK_WITH_CALL no longer generated with
single tuple or dict.
* Restored more informative error messages for incorrect var-positional and
var-keyword arguments.
* Removed code duplications in _PyEval_EvalCodeWithName().
* Removed redundant runtime checks and parameters in _PyStack_AsDict().
* Added a workaround and enabled previously disabled test in test_traceback.
* Removed dead code from the dis module.
Issue #27213: Rework CALL_FUNCTION* opcodes to produce shorter and more
efficient bytecode:
* CALL_FUNCTION now only accepts position arguments
* CALL_FUNCTION_KW accepts position arguments and keyword arguments, but keys
of keyword arguments are packed into a constant tuple.
* CALL_FUNCTION_EX is the most generic, it expects a tuple and a dict for
positional and keyword arguments.
CALL_FUNCTION_VAR and CALL_FUNCTION_VAR_KW opcodes have been removed.
2 tests of test_traceback are currently broken: skip test, the issue #28050 was
created to track the issue.
Patch by Demur Rumed, design by Serhiy Storchaka, reviewed by Serhiy Storchaka
and Victor Stinner.
Issue #26107: The format of the co_lnotab attribute of code objects changes to
support negative line number delta.
Changes:
* assemble_lnotab(): if line number delta is less than -128 or greater than
127, emit multiple (offset_delta, lineno_delta) in co_lnotab
* update functions decoding co_lnotab to use signed 8-bit integers
- dis.findlinestarts()
- PyCode_Addr2Line()
- _PyCode_CheckLineNumber()
- frame_setlineno()
* update lnotab_notes.txt
* increase importlib MAGIC_NUMBER to 3361
* document the change in What's New in Python 3.6
* cleanup also PyCode_Optimize() to use better variable names
- confusing line_offset parameter -> first_line parameter
- systematically test and fix new file parameter
- remove redundant Bytecode.show_info() API
- rename Bytecode.display_code() to Bytecode.dis() and have it
return the multi-line string rather than printing it directly
- eliminated some not-so-helpful helpers from the bytecode_helper
test support module
Also fixed a longstanding defect (worked around in the test suite)
where lines emitted by the dis module could include trailing white
space. That no longer happens, allowing the formatting tests to be
simplified to use plain string comparisons.
* get_instructions generator
* ability to redirect output to a file
* Bytecode and Instruction abstractions
Patch by Nick Coghlan, Ryan Kelly and Thomas Kluyver.
Also removed im_class and renamed im_self to __self__ and im_func to __func__. im_class can be substituted with method.__self__.__class__.
I've also updated some parts of the documenation.
svn+ssh://pythondev@svn.python.org/python/branches/p3yk
........
r55797 | neal.norwitz | 2007-06-07 00:00:57 -0700 (Thu, 07 Jun 2007) | 3 lines
Get rid of some remnants of classic classes. types.ClassType == type.
Also get rid of almost all uses of the types module and use the builtin name.
........
r55798 | neal.norwitz | 2007-06-07 00:12:36 -0700 (Thu, 07 Jun 2007) | 1 line
Remove a use of types, verify commit hook works
........
r55809 | guido.van.rossum | 2007-06-07 11:11:29 -0700 (Thu, 07 Jun 2007) | 2 lines
Fix syntax error introduced by Neal in last checkin.
........
svn+ssh://pythondev@svn.python.org/python/branches/p3yk
................
r55636 | neal.norwitz | 2007-05-29 00:06:39 -0700 (Tue, 29 May 2007) | 149 lines
Merged revisions 55506-55635 via svnmerge from
svn+ssh://pythondev@svn.python.org/python/trunk
........
r55507 | georg.brandl | 2007-05-22 07:28:17 -0700 (Tue, 22 May 2007) | 2 lines
Remove the "panel" module doc file which has been ignored since 1994.
........
r55522 | mark.hammond | 2007-05-22 19:04:28 -0700 (Tue, 22 May 2007) | 4 lines
Remove definition of PY_UNICODE_TYPE from pyconfig.h, allowing the
definition in unicodeobject.h to be used, giving us the desired
wchar_t in place of 'unsigned short'. As discussed on python-dev.
........
r55525 | neal.norwitz | 2007-05-22 23:35:32 -0700 (Tue, 22 May 2007) | 6 lines
Add -3 option to the interpreter to warn about features that are
deprecated and will be changed/removed in Python 3.0.
This patch is mostly from Anthony. I tweaked some format and added
a little doc.
........
r55527 | neal.norwitz | 2007-05-22 23:57:35 -0700 (Tue, 22 May 2007) | 1 line
Whitespace cleanup
........
r55528 | neal.norwitz | 2007-05-22 23:58:36 -0700 (Tue, 22 May 2007) | 1 line
Add a bunch more deprecation warnings for builtins that are going away in 3.0
........
r55549 | georg.brandl | 2007-05-24 09:49:29 -0700 (Thu, 24 May 2007) | 2 lines
shlex.split() now has an optional "posix" parameter.
........
r55550 | georg.brandl | 2007-05-24 10:33:33 -0700 (Thu, 24 May 2007) | 2 lines
Fix parameter passing.
........
r55555 | facundo.batista | 2007-05-24 10:50:54 -0700 (Thu, 24 May 2007) | 6 lines
Added an optional timeout parameter to urllib.ftpwrapper, with tests
(for this and a basic one, because there weren't any). Changed also
NEWS, but didn't find documentation for this function, assumed it
wasn't public...
........
r55563 | facundo.batista | 2007-05-24 13:01:59 -0700 (Thu, 24 May 2007) | 4 lines
Removed the .recv() in the test, is not necessary, and was
causing problems that didn't have anything to do with was
actually being tested...
........
r55564 | facundo.batista | 2007-05-24 13:51:19 -0700 (Thu, 24 May 2007) | 5 lines
Let's see if reading exactly what is written allow this live
test to pass (now I know why there were so few tests in ftp,
http, etc, :( ).
........
r55567 | facundo.batista | 2007-05-24 20:10:28 -0700 (Thu, 24 May 2007) | 4 lines
Trying to make the tests work in Windows and Solaris, everywhere
else just works
........
r55568 | facundo.batista | 2007-05-24 20:47:19 -0700 (Thu, 24 May 2007) | 4 lines
Fixing stupid error, and introducing a sleep, to see if the
other thread is awakened and finish sending data.
........
r55569 | facundo.batista | 2007-05-24 21:20:22 -0700 (Thu, 24 May 2007) | 4 lines
Commenting out the tests until find out who can test them in
one of the problematic enviroments.
........
r55570 | neal.norwitz | 2007-05-24 22:13:40 -0700 (Thu, 24 May 2007) | 2 lines
Get test passing again by commenting out the reference to the test class.
........
r55575 | vinay.sajip | 2007-05-25 00:05:59 -0700 (Fri, 25 May 2007) | 1 line
Updated docstring for SysLogHandler (#1720726).
........
r55576 | vinay.sajip | 2007-05-25 00:06:55 -0700 (Fri, 25 May 2007) | 1 line
Updated documentation for SysLogHandler (#1720726).
........
r55592 | brett.cannon | 2007-05-25 13:17:15 -0700 (Fri, 25 May 2007) | 3 lines
Remove direct call's to file's constructor and replace them with calls to
open() as ths is considered best practice.
........
r55601 | kristjan.jonsson | 2007-05-26 12:19:50 -0700 (Sat, 26 May 2007) | 1 line
Remove the rgbimgmodule from PCBuild8
........
r55602 | kristjan.jonsson | 2007-05-26 12:31:39 -0700 (Sat, 26 May 2007) | 1 line
Include <windows.h> after python.h, so that WINNT is properly set before windows.h is included. Fixes warnings in PC builds.
........
r55603 | walter.doerwald | 2007-05-26 14:04:13 -0700 (Sat, 26 May 2007) | 2 lines
Fix typo.
........
r55604 | peter.astrand | 2007-05-26 15:18:20 -0700 (Sat, 26 May 2007) | 1 line
Applied patch 1669481, slightly modified: Support close_fds on Win32
........
r55606 | neal.norwitz | 2007-05-26 21:08:54 -0700 (Sat, 26 May 2007) | 2 lines
Add the new function object attribute names from py3k.
........
r55617 | lars.gustaebel | 2007-05-27 12:49:30 -0700 (Sun, 27 May 2007) | 20 lines
Added errors argument to TarFile class that allows the user to
specify an error handling scheme for character conversion. Additional
scheme "utf-8" in read mode. Unicode input filenames are now
supported by design. The values of the pax_headers dictionary are now
limited to unicode objects.
Fixed: The prefix field is no longer used in PAX_FORMAT (in
conformance with POSIX).
Fixed: In read mode use a possible pax header size field.
Fixed: Strip trailing slashes from pax header name values.
Fixed: Give values in user-specified pax_headers precedence when
writing.
Added unicode tests. Added pax/regtype4 member to testtar.tar all
possible number fields in a pax header.
Added two chapters to the documentation about the different formats
tarfile.py supports and how unicode issues are handled.
........
r55618 | raymond.hettinger | 2007-05-27 22:23:22 -0700 (Sun, 27 May 2007) | 1 line
Explain when groupby() issues a new group.
........
r55634 | martin.v.loewis | 2007-05-28 21:01:29 -0700 (Mon, 28 May 2007) | 2 lines
Test pre-commit hook for a link to a .py file.
........
r55635 | martin.v.loewis | 2007-05-28 21:02:03 -0700 (Mon, 28 May 2007) | 2 lines
Revert 55634.
........
................
r55639 | neal.norwitz | 2007-05-29 00:58:11 -0700 (Tue, 29 May 2007) | 1 line
Remove sys.exc_{type,exc_value,exc_traceback}
................
r55641 | neal.norwitz | 2007-05-29 01:03:50 -0700 (Tue, 29 May 2007) | 1 line
Missed one sys.exc_type. I wonder why exc_{value,traceback} were already gone
................
r55642 | neal.norwitz | 2007-05-29 01:08:33 -0700 (Tue, 29 May 2007) | 1 line
Missed more doc for sys.exc_* attrs.
................
r55643 | neal.norwitz | 2007-05-29 01:18:19 -0700 (Tue, 29 May 2007) | 1 line
Remove sys.exc_clear()
................
r55665 | guido.van.rossum | 2007-05-29 19:45:43 -0700 (Tue, 29 May 2007) | 4 lines
Make None, True, False keywords.
We can now also delete all the other places that explicitly forbid
assignment to None, but I'm not going to bother right now.
................
r55666 | guido.van.rossum | 2007-05-29 20:01:51 -0700 (Tue, 29 May 2007) | 3 lines
Found another place that needs check for forbidden names.
Fixed test_syntax.py accordingly (it helped me find that one).
................
r55668 | guido.van.rossum | 2007-05-29 20:41:48 -0700 (Tue, 29 May 2007) | 2 lines
Mark None, True, False as keywords.
................
r55673 | neal.norwitz | 2007-05-29 23:28:25 -0700 (Tue, 29 May 2007) | 3 lines
Get the dis module working on modules again after changing dicts
to not return lists and also new-style classes. Add a test.
................
r55674 | neal.norwitz | 2007-05-29 23:35:45 -0700 (Tue, 29 May 2007) | 1 line
Umm, it helps to add the module that the test uses
................
r55675 | neal.norwitz | 2007-05-29 23:53:05 -0700 (Tue, 29 May 2007) | 4 lines
Try to fix up all the other places that were assigning to True/False.
There's at least one more problem in test.test_xmlrpc. I have other
changes in that file and that should be fixed soon (I hope).
................
r55679 | neal.norwitz | 2007-05-30 00:31:55 -0700 (Wed, 30 May 2007) | 1 line
Fix up another place that was assigning to True/False.
................
r55688 | brett.cannon | 2007-05-30 14:19:47 -0700 (Wed, 30 May 2007) | 2 lines
Ditch MimeWriter.
................
r55692 | brett.cannon | 2007-05-30 14:52:00 -0700 (Wed, 30 May 2007) | 2 lines
Remove the mimify module.
................
r55707 | guido.van.rossum | 2007-05-31 05:08:45 -0700 (Thu, 31 May 2007) | 2 lines
Backport the addition of show_code() to dis.py -- it's too handy.
................
r55708 | guido.van.rossum | 2007-05-31 06:22:57 -0700 (Thu, 31 May 2007) | 7 lines
Fix a fairly long-standing bug in the check for assignment to None (and other
keywords, these days). In 2.5, you could write foo(None=1) without getting
a SyntaxError (although foo()'s definition would have to use **kwds to avoid
getting a runtime error complaining about an unknown keyword of course).
This ought to be backported to 2.5.2 or at least 2.6.
................
r55724 | brett.cannon | 2007-05-31 19:32:41 -0700 (Thu, 31 May 2007) | 2 lines
Remove the cfmfile.
................
r55727 | neal.norwitz | 2007-05-31 22:19:44 -0700 (Thu, 31 May 2007) | 1 line
Remove reload() builtin.
................
r55729 | neal.norwitz | 2007-05-31 22:51:30 -0700 (Thu, 31 May 2007) | 59 lines
Merged revisions 55636-55728 via svnmerge from
svn+ssh://pythondev@svn.python.org/python/trunk
........
r55637 | georg.brandl | 2007-05-29 00:16:47 -0700 (Tue, 29 May 2007) | 2 lines
Fix rst markup.
........
r55638 | neal.norwitz | 2007-05-29 00:51:39 -0700 (Tue, 29 May 2007) | 1 line
Fix typo in doc
........
r55671 | neal.norwitz | 2007-05-29 21:53:41 -0700 (Tue, 29 May 2007) | 1 line
Fix indentation (whitespace only).
........
r55676 | thomas.heller | 2007-05-29 23:58:30 -0700 (Tue, 29 May 2007) | 1 line
Fix compiler warnings.
........
r55677 | thomas.heller | 2007-05-30 00:01:25 -0700 (Wed, 30 May 2007) | 2 lines
Correct the name of a field in the WIN32_FIND_DATAA and WIN32_FIND_DATAW structures.
Closes bug #1726026.
........
r55686 | brett.cannon | 2007-05-30 13:46:26 -0700 (Wed, 30 May 2007) | 2 lines
Have MimeWriter raise a DeprecationWarning as per PEP 4 and its documentation.
........
r55690 | brett.cannon | 2007-05-30 14:48:58 -0700 (Wed, 30 May 2007) | 3 lines
Have mimify raise a DeprecationWarning. The docs and PEP 4 have listed the
module as deprecated for a while.
........
r55696 | brett.cannon | 2007-05-30 15:24:28 -0700 (Wed, 30 May 2007) | 2 lines
Have md5 raise a DeprecationWarning as per PEP 4.
........
r55705 | neal.norwitz | 2007-05-30 21:14:22 -0700 (Wed, 30 May 2007) | 1 line
Add some spaces in the example code.
........
r55716 | brett.cannon | 2007-05-31 12:20:00 -0700 (Thu, 31 May 2007) | 2 lines
Have the sha module raise a DeprecationWarning as specified in PEP 4.
........
r55719 | brett.cannon | 2007-05-31 12:40:42 -0700 (Thu, 31 May 2007) | 2 lines
Cause buildtools to raise a DeprecationWarning.
........
r55721 | brett.cannon | 2007-05-31 13:01:11 -0700 (Thu, 31 May 2007) | 2 lines
Have cfmfile raise a DeprecationWarning as per PEP 4.
........
r55726 | neal.norwitz | 2007-05-31 21:56:47 -0700 (Thu, 31 May 2007) | 1 line
Mail if there is an installation failure.
........
................
r55730 | neal.norwitz | 2007-05-31 23:22:07 -0700 (Thu, 31 May 2007) | 2 lines
Remove the code that was missed in rev 55303.
................
r55738 | neal.norwitz | 2007-06-01 19:10:43 -0700 (Fri, 01 Jun 2007) | 1 line
Fix doc breakage
................
r55741 | neal.norwitz | 2007-06-02 00:41:58 -0700 (Sat, 02 Jun 2007) | 1 line
Remove timing module (plus some remnants of other modules).
................
r55742 | neal.norwitz | 2007-06-02 00:51:44 -0700 (Sat, 02 Jun 2007) | 1 line
Remove posixfile module (plus some remnants of other modules).
................
r55744 | neal.norwitz | 2007-06-02 10:18:56 -0700 (Sat, 02 Jun 2007) | 1 line
Fix doc breakage.
................
r55745 | neal.norwitz | 2007-06-02 11:32:16 -0700 (Sat, 02 Jun 2007) | 1 line
Make a whatsnew 3.0 template.
................
r55754 | neal.norwitz | 2007-06-03 23:24:18 -0700 (Sun, 03 Jun 2007) | 1 line
SF #1730441, os._execvpe raises UnboundLocal due to new try/except semantics
................
r55755 | neal.norwitz | 2007-06-03 23:26:00 -0700 (Sun, 03 Jun 2007) | 1 line
Get rid of extra whitespace
................
r55794 | guido.van.rossum | 2007-06-06 15:29:22 -0700 (Wed, 06 Jun 2007) | 3 lines
Make this compile in GCC 2.96, which does not allow interspersing
declarations and code.
................
There's one major and one minor category still unfixed:
doctests are the major category (and I hope to be able to augment the
refactoring tool to refactor bona fide doctests soon);
other code generating print statements in strings is the minor category.
(Oh, and I don't know if the compiler package works.)
Use a slightly different strategy to determine when not to call the line
trace function. This removes the need for the RETURN_NONE opcode, so
that's gone again. Update docs and comments to match.
Thanks to Neal and Armin!
Also add a test suite. This should have come with the original patch...
[ 587993 ] SET_LINENO killer
Remove SET_LINENO. Tracing is now supported by inspecting co_lnotab.
Many sundry changes to document and adapt to this change.
new slot tp_iter in type object, plus new flag Py_TPFLAGS_HAVE_ITER
new C API PyObject_GetIter(), calls tp_iter
new builtin iter(), with two forms: iter(obj), and iter(function, sentinel)
new internal object types iterobject and calliterobject
new exception StopIteration
new opcodes for "for" loops, GET_ITER and FOR_ITER (also supported by dis.py)
new magic number for .pyc files
new special method for instances: __iter__() returns an iterator
iteration over dictionaries: "for x in dict" iterates over the keys
iteration over files: "for x in file" iterates over lines
TODO:
documentation
test suite
decide whether to use a different way to spell iter(function, sentinal)
decide whether "for key in dict" is a good idea
use iterators in map/filter/reduce, min/max, and elsewhere (in/not in?)
speed tuning (make next() a slot tp_next???)
added test script and expected output file as well
this closes patch 103297.
__all__ attributes will be added to other modules without first submitting
a patch, just adding the necessary line to the test script to verify
more-or-less correct implementation.
rfc822 (Addresslist) modules. Also a preliminary testcase for augmented
assignment, which should actually be merged with the test_class testcase I
added last week.
Add the EXTENDED_ARG opcode to the virtual machine, allowing 32-bit
arguments to opcodes instead of being forced to stick to the 16-bit
limit. This is especially useful for machine-generated code, which
can be too long for the SET_LINENO parameter to fit into 16 bits.
This closes the implementation portion of SourceForge patch #100893.
name as n'. By doing some twists and turns, "as" is not a reserved word.
There is a slight change in semantics for 'from module import name' (it will
now honour the 'global' keyword) but only in cases that are explicitly
undocumented.
did the same anyway.
I'm not sure what to do with Tools/compiler/compiler/* -- that isn't part of
distutils, is it ? Should it try to be compatible with old bytecode version ?
executive summary:
Instead of typing 'apply(f, args, kwargs)' you can type 'f(*arg, **kwargs)'.
Some file-by-file details follow.
Grammar/Grammar:
simplify varargslist, replacing '*' '*' with '**'
add * & ** options to arglist
Include/opcode.h & Lib/dis.py:
define three new opcodes
CALL_FUNCTION_VAR
CALL_FUNCTION_KW
CALL_FUNCTION_VAR_KW
Python/ceval.c:
extend TypeError "keyword parameter redefined" message to include
the name of the offending keyword
reindent CALL_FUNCTION using four spaces
add handling of sequences and dictionaries using extend calls
fix function import_from to use PyErr_Format
Mark Shannon
Brandt Bucher
Saul Shanabrook
Nikita Sobolev
Brandt Bucher
Irit Katriel
Batuhan Taskaya
Eric Snow
Pablo Galindo
Yurii Karabas
Konge
T. Wouters
Serhiy Storchaka
syncosmic
Victor Stinner
Yury Selivanov
Raymond Hettinger
Yury Selivanov
Nick Coghlan
Ezio Melotti
Alexander Belopolsky
Benjamin Peterson
Alexandre Vassalotti
Christian Heimes
Guido van Rossum
Collin Winter
Neal Norwitz
Walter Dörwald
Armin Rigo
Tim Peters
Skip Montanaro
Michael W. Hudson
Jeremy Hylton
Eric S. Raymond
Fred Drake
Barry Warsaw