Fix overflow bug in ldexp(x, exp). The 'exp' argument maps to a C int for the
math library call [double ldexp(double, int)], however the 'd'
PyArg_ParseTuple formatter was used to yield a double, which was subsequently
cast to an int. This could overflow.
[GvR: mysteriously, on Solaris 2.7, ldexp(1, 2147483647) returns Inf
while ldexp(1, 2147483646) raises OverflowError; this seems a bug in
the math library (it also takes a real long time to compute the
Inf outcome). Does this point to a bug in the CHECK() macro? It
should have discovered that the result was outside the HUGE_VAL range.]
The following modules are specifically excluded in the Win64 build:
audioop, binascii, imageop, rgbimg. They are advertised as heavily 32-bit
dependent. [They should probably be fixed! --GvR]
Changes to PC\config.[hc] for Win64. MSVC defines _WINxx to differentiate the
various windows platforms. Python's MS_WINxx are keyed off of these. Note
that _WIN32 (and hence MS_WIN32 in Python) are defined on Win32 *and* on
Win64. This is for compatibility reasons. The idea is that the common case is
that code specific to Win32 will also work on Win64 rather than being
specific to Win32 (i.e. there is more the same than different in WIn32 and
Win64).
The following modules are specifically excluded in the Win64 build:
audioop, binascii, imageop, rgbimg. They are advertised as heavily 32-bit
dependent. [They should probably be fixed! --GvR]
The patch to config.h looks big but it really is not. These are the effective
changes:
- MS_WINxx are keyed off _WINxx
- SIZEOF_VOID_P is set to 8 for Win64
- COMPILER string is changed appropriately for Win64
Fix the string methods that implement slice-like semantics with
optional args (count, find, endswith, etc.) to properly handle
indeces outside [INT_MIN, INT_MAX]. Previously the "i" formatter
for PyArg_ParseTuple was used to get the indices. These could overflow.
This patch changes the string methods to use the "O&" formatter with
the slice_index() function from ceval.c which is used to do the same
job for Python code slices (e.g. 'abcabcabc'[0:1000000000L]). slice_index()
is renamed _PyEval_SliceIndex() and is now exported. As well, the return
values for success/fail were changed to make slice_index directly
usable as required by the "O&" formatter.
[GvR: shouldn't a similar patch be applied to unicodeobject.c?]
Changes the 'b', 'h', and 'i' formatters in PyArg_ParseTuple to raise an
Overflow exception if they overflow (previously they just silently
overflowed).
Changes by Guido: always accept values [0..255] (in addition to
[CHAR_MIN..CHAR_MAX]) for 'b' format; changed some spaces into tabs in
other code.
gave bogus results for chars in the range 128-255, because their
implementation was using signed characters. Fixed this by using
unsigned character pointers (as opposed to using Py_CHARMASK()).
weird errors. (E.g. see thread "weird bug in test_winreg" in python-dev.)
Since it's actually useful to be able to re-run an individual test
after running test.autotest, we keep the unloading code, but only for
modules whose full name starts with "test.".
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
instead. This seems more robust than returning an Unicode string with
some unconverted charcters in it.
This still doesn't support getting truly binary data out of Tcl, since
we look for the trailing null byte; but the old (pre-Unicode) code did
this too, so apparently there's no need. (Plus, I really don't feel
like finding out how Tcl deals with this in each version.)
1. In Tcl 8.2 and later, use Tcl_NewUnicodeObj() when passing a Python
Unicode object rather than going through UTF-8. (This function
doesn't exist in Tcl 8.1, so there the original UTF-8 code is still
used; in Tcl 8.0 there is no support for Unicode.) This assumes that
Tcl_UniChar is the same thing as Py_UNICODE; a run-time error is
issued if this is not the case.
2. In Tcl 8.1 and later (i.e., whenever Tcl supports Unicode), when a
string returned from Tcl contains bytes with the top bit set, we
assume it is encoded in UTF-8, and decode it into a Unicode string
object.
Notes:
- Passing Unicode strings to Tcl 8.0 does not do the right thing; this
isn't worth fixing.
- When passing an 8-bit string to Tcl 8.1 or later that has bytes with
the top bit set, Tcl tries to interpret it as UTF-8; it seems to fall
back on Latin-1 for non-UTF-8 bytes. I'm not sure what to do about
this besides telling the user to disambiguate such strings by
converting them to Unicode (forcing the user to be explicit about the
encoding).
- Obviously it won't be possible to get binary data out of Tk this
way. Do we need that ability? How to do it?
For more comments, read the patches@python.org archives.
For documentation read the comments in mymalloc.h and objimpl.h.
(This is not exactly what Vladimir posted to the patches list; I've
made a few changes, and Vladimir sent me a fix in private email for a
problem that only occurs in debug mode. I'm also holding back on his
change to main.c, which seems unnecessary to me.)
Guido van Rossum
Jeremy Hylton
Jack Jansen
Greg Ward