The ob_sval member of a string object isn't necessarily aligned to better
than a native long, so the new "q" and "Q" struct codes can't get away w/
casting tricks on platforms where LONG_LONG requires stricter-than-long
alignment. After I thought of a few elaborate workarounds, Guido bashed
me over the head with the obvious memcpy approach, herewith implemented.
This completes the q/Q project.
longobject.c _PyLong_AsByteArray: The original code had a gross bug:
the most-significant Python digit doesn't necessarily have SHIFT
significant bits, and you really need to count how many copies of the sign
bit it has else spurious overflow errors result.
test_struct.py: This now does exhaustive std q/Q testing at, and on both
sides of, all relevant power-of-2 boundaries, both positive and negative.
NEWS: Added brief dict news while I was at it.
native mode, and only when config #defines HAVE_LONG_LONG. Standard mode
will eventually treat them as 8-byte ints across all platforms, but that
likely requires a new set of routines in longobject.c first (while
sizeof(long) >= 4 is guaranteed by C, there's nothing in C we can rely
on x-platform to hold 8 bytes of int, so we'll have to roll our own;
I'm thinking of a simple pair of conversion functions, Python long
to/from sized vector of unsigned bytes; that may be useful for GMP
conversions too; std q/Q would call them with size fixed at 8).
test_struct.py: In addition to adding some native-mode 'q' and 'Q' tests,
got rid of unused code, and repaired a non-portable assumption about
native sizeof(short) (it isn't 2 on some Cray boxes).
libstruct.tex: In addition to adding a bit of 'q'/'Q' docs (more needed
later), removed an erroneous footnote about 'I' behavior.
where sizeof(long)==8. This *was* broken on boxes where signed right
shifts didn't sign-extend, but not elsewhere. Unfortunately, apart
from the Cray T3E I don't know of such a box, and Guido has so far
refused to buy me any Cray machines for home Python testing <wink>.
More immediately interesting would be if someone could please test
this on *any* sizeof(long)==8 box, to make sure I didn't break it.
Add definitions of INT_MAX and LONG_MAX to pyport.h.
Remove includes of limits.h and conditional definitions of INT_MAX
and LONG_MAX elsewhere.
This closes SourceForge patch #101659 and bug #115323.
and a couple of functions that were missed in the previous batches. Not
terribly tested, but very carefully scrutinized, three times.
All these were found by the little findkrc.py that I posted to python-dev,
which means there might be more lurking. Cases such as this:
long
func(a, b)
long a;
long b; /* flagword */
{
and other cases where the last ; in the argument list isn't followed by a
newline and an opening curly bracket. Regexps to catch all are welcome, of
course ;)
(1) Use PyErr_NewException("module.class", NULL, NULL) to create the
exception object.
(2) Remove all calls to Py_FatalError(); instead, return or
ignore the errors -- the import code now checks PyErr_Occurred()
after calling a module's init function, so it's no longer a
fatal error for the initialization to fail.
Also did some small cleanups, e.g. removed unnecessary test for
"already initialized" from initfpectl(), and unified
initposix()/initnt().
I haven't checked this very thoroughly, so while the changes are
pretty trivial -- beware of untested code!
Added 'p' format character for Pascal string (i.e. leading length
byte). This uses the count prefix line 's' does, except that the
count includes the length byte; i.e. '10p' takes 10 bytes packed but
has space for a length byte and 9 data bytes.
int/long types, and use the new PyLong_FromUnsignedLong() and
PyLong_AsUnsignedLong() interfaces instead.
Semantic change: the 'I' format will now always return a long int.
-- The whole implementation is now more table-driven.
-- Unsigned integers. Format characters 'B', 'H', 'I' and 'L'
mean unsigned byte, short, int and long. For 'I' and 'L', the return
value is a Python long integer if a Python plain integer can't
represent the required range (note: this is dependent on the size of
the relevant C types only, not of the sign of the actual value).
-- A new format character 's' packs/unpacks a string. When given a
count prefix, this is the size of the string, not a repeat count like
for the other format characters; e.g. '10s' means a single 10-byte
string, while '10c' means 10 characters. For packing, the string is
truncated or padded with null bytes as appropriate to make it fit.
For unpacking, the resulting string always has exactly the specified
number of bytes. As a special case, '0s' means a single, empty
string (while '0c' means 0 characters).
-- Various byte order options. The first character of the format
string determines the byte order, size and alignment, as follows:
First character Byte order size and alignment
'@' native native
'=' native standard
'<' little-endian standard
'>' big-endian standard
'!' network (= big-endian) standard
If the first character is not one of these, '@' is assumed.
Native byte order is big-endian or little-endian, depending on the
host system (e.g. Motorola and Sun are big-endian; Intel and DEC are
little-endian).
Native size and alignment are determined using the C compiler's sizeof
expression. This is always combined with native byte order.
Standard size and alignment are as follows: no alignment is required
for any type (so you have to use pad bytes); short is 2 bytes; int and
long are 4 bytes. In this mode, there is no support for float and
double.
Note the difference between '@' and '=': both use native byte order,
but the size and alignment of the latter is standardized.
The form '!' is available for those poor souls who can't remember
whether network byte order is big-endian or little-endian.
There is no way to indicate non-native byte order (i.e. force
byte-swapping); use the appropriate choice of '<' or '>'.
