and reported to python-dev: because we were calling dict_resize() in
PyDict_Next(), and because GC's dict_traverse() uses PyDict_Next(),
and because PyTuple_New() can cause GC, and because dict_items() calls
PyTuple_New(), it was possible for dict_items() to have the dict
resized right under its nose.
The solution is convoluted, and touches several places: keys(),
values(), items(), popitem(), PyDict_Next(), and PyDict_SetItem().
There are two parts to it. First, we no longer call dict_resize() in
PyDict_Next(), which seems to solve the immediate problem. But then
PyDict_SetItem() must have a different policy about when *it* calls
dict_resize(), because we want to guarantee (e.g. for an algorithm
that Jeremy uses in the compiler) that you can loop over a dict using
PyDict_Next() and make changes to the dict as long as those changes
are only value replacements for existing keys using PyDict_SetItem().
This is done by resizing *after* the insertion instead of before, and
by remembering the size before we insert the item, and if the size is
still the same, we don't bother to even check if we might need to
resize. An additional detail is that if the dict starts out empty, we
must still resize it before the insertion.
That was the first part. :-)
The second part is to make keys(), values(), items(), and popitem()
safe against side effects on the dict caused by allocations, under the
assumption that if the GC can cause arbitrary Python code to run, it
can cause other threads to run, and it's not inconceivable that our
dict could be resized -- it would be insane to write code that relies
on this, but not all code is sane.
Now, I have this nagging feeling that the loops in lookdict probably
are blissfully assuming that doing a simple key comparison does not
change the dict's size. This is not necessarily true (the keys could
be class instances after all). But that's a battle for another day.
later. This assumes that zlib.h has a line of the form
#define ZLIB_VERSION "1.1.3"
This solves the problem where a zlib installation is found but it is
an older version -- this would break the build, while a better
solution is to simply ignore that zlib installation.
cut-and-paste copy of the seek() method on the _Subfile class, but it
didn't make one bit of sense: it sets self.pos, which is not used in
this class or its subclasses, and it uses self.start and self.stop,
which aren't defined on this class or its subclasses. This is purely
my own fault -- I added this in rev 1.4 and apparently never tried to
use it. Since it's not documented, and of very questionable use given
that there's no tell(), I'm ripping it out.
This resolves SF bug 416199 by Andrew Dalke: mailbox.py seek problems.
ZipFile.close() method that should be part of the preceding 'if'
block. On some platforms (Mark noticed this on FreeBSD 4.2) doing a
flush() on a file open for reading is not allowed.
always:
- #undef HAVE_CONFIG_H (because otherwise chardefs.h tries to include
strings.h)
- #include readline.h and history.h
and we never declare any readline function prototypes ourselves.
This makes it compile with readline 4.2, albeit with a few warnings.
Some of the remaining warnings are about completion_matches(), which
is renamed to rl_completion_matches().
I've tested it with various other versions, from 2.0 up, and they all
seem to work (some with warnings) -- but only on Red Hat Linux 6.2.
Fixing the warnings for readline 4.2 would break compatibility with
3.0 (and maybe even earlier versions), and readline doesn't seem to
have a way to test for its version at compile time, so I'd rather
leave the warnings in than break compilation with older versions.
problem reported by Neil Schemenauer on python-dev on 4/12/01, wth
subject "Problem with SSL and socketmodule on Debian Potato?".
It's tentative because Moshe objected, but Martin rebutted, and Moshe
seems unavailable for comments.
(Note that with OpenSSL 0.9.6a, I get a lot of compilation warnings
for socketmodule.c -- I'm assuming I can safely ignore these until 2.1
is released.)
before calling any callbacks. This is important
since the callback objects only look at themselves
to determine that they are invalide. This change
avoids a segfault when callbacks use a different
reference to an object in the process of being
deallocated.
This fixes SF bug #415660.
(with modification of existing dict elements!).
This is part of SF patch #409864: lazy fix for Pings bizarre scoping
crash.
The adaptation I made to Michael's patch was to change the error
handling to avoid masking other errors (moving the specific error
message to inside test_dict_inner()), and to insert a test for
dict==NULL at the start.
now raises NameError instead of UnboundLocalError, because the var in
question is definitely not local. (This affects test_scope.py)
Also update the recent fix by Ping using get_func_name(). Replace
tests of get_func_name() return value with call to get_func_desc() to
match all the other uses.
Calling an unbound method on a C extension class without providing
an instance can yield a segfault. Try "Exception.__init__()" or
"ValueError.__init__()".
This is a simple fix. The error-reporting bits in call_method
mistakenly treat the misleadingly-named variable "func" as a
function, when in fact it is a method.
If we let get_func_name take care of the work, all is fine.
- Removed the subsection numbering in section B (each time a new
license is inserted in the front, the others have to be renumbered).
- Changed the words in the intro to avoid implying that 1.6.1 is
GPL-compatible.
Guido van Rossum
Fredrik Lundh
Steve Purcell
Fred Drake
Eric S. Raymond
cvs2svn
Jeremy Hylton
Ka-Ping Yee