]> granicus.if.org Git - python/log
python
24 years agoAdded a WITHOUT_FRAMEWORKS define to all the config files, so that on MacOS<=9 compil...
Jack Jansen [Mon, 14 May 2001 15:00:38 +0000 (15:00 +0000)]
Added a WITHOUT_FRAMEWORKS define to all the config files, so that on MacOS<=9 compiles use Universal Headers, not Carbon/Carbon.h.

24 years agoFix a typo, consistently spell ASCII in all caps, and insert blank
Guido van Rossum [Mon, 14 May 2001 13:53:38 +0000 (13:53 +0000)]
Fix a typo, consistently spell ASCII in all caps, and insert blank
lines between paragraphs in Mark Hammond's news item about the default
encoding in posixmodule.  Resist the temptation to reflow paragraphs.

24 years agoFix the Py_FileSystemDefaultEncoding checkin - declare the variable in a fileobject...
Mark Hammond [Mon, 14 May 2001 12:17:34 +0000 (12:17 +0000)]
Fix the Py_FileSystemDefaultEncoding checkin - declare the variable in a fileobject.h, and initialize it in bltinmodule.

24 years agoFix the .find() method for memory maps.
Greg Stein [Mon, 14 May 2001 09:32:26 +0000 (09:32 +0000)]
Fix the .find() method for memory maps.

1) it didn't obey the "start" parameter (and when it does, we must validate
   the value)
2) the return value needs to be an absolute index, rather than relative to
   some arbitrary point in the file

(checking CVS, it appears this method never worked; these changes bring it
 into line with typical .find() behavior)

24 years agoSF bug[ #423781: pprint.isrecursive() broken.
Tim Peters [Mon, 14 May 2001 07:05:58 +0000 (07:05 +0000)]
SF bug[ #423781:  pprint.isrecursive() broken.

24 years agoAdd mention of the default file system encoding for Windows.
Mark Hammond [Mon, 14 May 2001 03:09:36 +0000 (03:09 +0000)]
Add mention of the default file system encoding for Windows.

24 years agoA disgusting "fix" for the test___all__ failure under Windows.
Tim Peters [Sun, 13 May 2001 09:01:06 +0000 (09:01 +0000)]
A disgusting "fix" for the test___all__ failure under Windows.

24 years agoAdd support for Windows using "mbcs" as the default Unicode encoding when dealing...
Mark Hammond [Sun, 13 May 2001 08:04:26 +0000 (08:04 +0000)]
Add support for Windows using "mbcs" as the default Unicode encoding when dealing with the file system.  As discussed on python-dev and in patch 410465.

24 years agoAggressive reordering of dict comparisons. In case of collision, it stands
Tim Peters [Sun, 13 May 2001 06:43:53 +0000 (06:43 +0000)]
Aggressive reordering of dict comparisons.  In case of collision, it stands
to reason that me_key is much more likely to match the key we're looking
for than to match dummy, and if the key is absent me_key is much more
likely to be NULL than dummy:  most dicts don't even have a dummy entry.
Running instrumented dict code over the test suite and some apps confirmed
that matching dummy was 200-300x less frequent than matching key in
practice.  So this reorders the tests to try the common case first.
It can lose if a large dict with many collisions is mostly deleted, not
resized, and then frequently searched, but that's hardly a case we
should be favoring.

24 years agoGet rid of the superstitious "~" in dict hashing's "i = (~hash) & mask".
Tim Peters [Sun, 13 May 2001 00:19:31 +0000 (00:19 +0000)]
Get rid of the superstitious "~" in dict hashing's "i = (~hash) & mask".
The comment following used to say:
/* We use ~hash instead of hash, as degenerate hash functions, such
   as for ints <sigh>, can have lots of leading zeros. It's not
   really a performance risk, but better safe than sorry.
   12-Dec-00 tim:  so ~hash produces lots of leading ones instead --
   what's the gain? */
That is, there was never a good reason for doing it.  And to the contrary,
as explained on Python-Dev last December, it tended to make the *sum*
(i + incr) & mask (which is the first table index examined in case of
collison) the same "too often" across distinct hashes.

Changing to the simpler "i = hash & mask" reduced the number of string-dict
collisions (== # number of times we go around the lookup for-loop) from about
6 million to 5 million during a full run of the test suite (these are
approximate because the test suite does some random stuff from run to run).
The number of collisions in non-string dicts also decreased, but not as
dramatically.

Note that this may, for a given dict, change the order (wrt previous
releases) of entries exposed by .keys(), .values() and .items().  A number
of std tests suffered bogus failures as a result.  For dicts keyed by
small ints, or (less so) by characters, the order is much more likely to be
in increasing order of key now; e.g.,

>>> d = {}
>>> for i in range(10):
...    d[i] = i
...
>>> d
{0: 0, 1: 1, 2: 2, 3: 3, 4: 4, 5: 5, 6: 6, 7: 7, 8: 8, 9: 9}
>>>

Unfortunately. people may latch on to that in small examples and draw a
bogus conclusion.

test_support.py
    Moved test_extcall's sortdict() into test_support, made it stronger,
    and imported sortdict into other std tests that needed it.
test_unicode.py
    Excluced cp875 from the "roundtrip over range(128)" test, because
    cp875 doesn't have a well-defined inverse for unicode("?", "cp875").
    See Python-Dev for excruciating details.
Cookie.py
    Chaged various output functions to sort dicts before building
    strings from them.
test_extcall
    Fiddled the expected-result file.  This remains sensitive to native
    dict ordering, because, e.g., if there are multiple errors in a
    keyword-arg dict (and test_extcall sets up many cases like that), the
    specific error Python complains about first depends on native dict
    ordering.

24 years agoGot the first MacPython module working under MacOSX/MachO (gestalt). Main changes
Jack Jansen [Sat, 12 May 2001 22:46:35 +0000 (22:46 +0000)]
Got the first MacPython module working under MacOSX/MachO (gestalt). Main changes
are including Carbon/Carbon.h in stead of the old headers (unless WITHOUT_FRAMEWORKS
is defined, as it will be for classic MacPython) and selectively disabling all the
stuff that is unneeded in a unix-Python (event handling, etc).

24 years agoBe more sensible about when to use TARGET_API_MAC_OS8 in stead of !TARGET_API_MAC_CAR...
Jack Jansen [Sat, 12 May 2001 21:31:34 +0000 (21:31 +0000)]
Be more sensible about when to use TARGET_API_MAC_OS8 in stead of !TARGET_API_MAC_CARBON. This should greatly facilitate porting stuff to OSX in its MachO/BSD incarnation.

24 years agoAdded iterobject.c to the project. And trying my first checkin at the same time.
Jack Jansen [Sat, 12 May 2001 21:09:45 +0000 (21:09 +0000)]
Added iterobject.c to the project. And trying my first checkin at the same time.

24 years agoRepair "module has no attribute xxx" error msg; bug introduced when
Tim Peters [Sat, 12 May 2001 20:24:22 +0000 (20:24 +0000)]
Repair "module has no attribute xxx" error msg; bug introduced when
switching from tp_getattr to tp_getattro.

24 years agoRefactored, with some future plans in mind.
Guido van Rossum [Sat, 12 May 2001 12:30:04 +0000 (12:30 +0000)]
Refactored, with some future plans in mind.

This now uses the new gotofileline() method defined in FileList.py.

24 years agoMove the action of loading the configuration to the IdleConf module
Guido van Rossum [Sat, 12 May 2001 12:18:10 +0000 (12:18 +0000)]
Move the action of loading the configuration to the IdleConf module
rather than the idle.py script.  This has advantages and
disadvantages; the biggest advantage being that we can more easily
have an alternative main program.

24 years agoDelete goodname() method, which is unused.
Guido van Rossum [Sat, 12 May 2001 12:11:36 +0000 (12:11 +0000)]
Delete goodname() method, which is unused.

Add gotofileline(), a convenience method which I intend to use in a
variant.

Rename test() to _test().

24 years agoVariant of patch #423262: Change module attribute get & set
Tim Peters [Fri, 11 May 2001 21:51:48 +0000 (21:51 +0000)]
Variant of patch #423262:  Change module attribute get & set
Allow module getattr and setattr to exploit string interning, via the
previously null module object tp_getattro and tp_setattro slots.   Yields
a very nice speedup for things like random.random and os.path etc.

24 years agoFix a minor style consistency issue.
Fred Drake [Fri, 11 May 2001 20:12:26 +0000 (20:12 +0000)]
Fix a minor style consistency issue.

When getting a string buffer for a string we just created, use
PyString_AS_STRING() instead of PyString_AsString() to avoid the
call overhead and extra type check.

24 years agoFix one bare except: clause.
Fred Drake [Fri, 11 May 2001 19:52:57 +0000 (19:52 +0000)]
Fix one bare except: clause.

24 years ago[].index() raises ValueError if the value is not in the list, so only
Fred Drake [Fri, 11 May 2001 19:52:03 +0000 (19:52 +0000)]
[].index() raises ValueError if the value is not in the list, so only
catch that instead of using a bare except clause.

24 years ago[].index() raises ValueError if the value is not in the list, so only
Fred Drake [Fri, 11 May 2001 19:44:55 +0000 (19:44 +0000)]
[].index() raises ValueError if the value is not in the list, so only
catch that instead of using a bare except clause.

24 years agoOnly catch NameError and TypeError when attempting to subclass an
Fred Drake [Fri, 11 May 2001 19:40:10 +0000 (19:40 +0000)]
Only catch NameError and TypeError when attempting to subclass an
exception (for compatibility with old versions of Python).

24 years agoRemove a bare try/except completely -- it just did not make sense!
Fred Drake [Fri, 11 May 2001 19:25:08 +0000 (19:25 +0000)]
Remove a bare try/except completely -- it just did not make sense!

Add a comment elsewhere making clear an assumption in the code.

24 years agoWhen guarding an import, only catch ImportError.
Fred Drake [Fri, 11 May 2001 19:21:41 +0000 (19:21 +0000)]
When guarding an import, only catch ImportError.

24 years agoClean up a bare except where we only expect to catch pcre.error.
Fred Drake [Fri, 11 May 2001 19:20:17 +0000 (19:20 +0000)]
Clean up a bare except where we only expect to catch pcre.error.

24 years agoClean up bare except where only IOError makes sense.
Fred Drake [Fri, 11 May 2001 19:15:28 +0000 (19:15 +0000)]
Clean up bare except where only IOError makes sense.

24 years agoClean up bare except: when determining whether a file is seekable.
Fred Drake [Fri, 11 May 2001 19:14:51 +0000 (19:14 +0000)]
Clean up bare except: when determining whether a file is seekable.

24 years agoOpening a file for reading can raise IOError, so only catch that.
Fred Drake [Fri, 11 May 2001 18:47:54 +0000 (18:47 +0000)]
Opening a file for reading can raise IOError, so only catch that.

24 years agoint() of a string is only expected to through ValueError, so do not use
Fred Drake [Fri, 11 May 2001 18:45:52 +0000 (18:45 +0000)]
int() of a string is only expected to through ValueError, so do not use
a bare except clause.

24 years ago<socket>.getsockopt() and <socket>.setsockopt() can only raise socket.error,
Fred Drake [Fri, 11 May 2001 18:28:54 +0000 (18:28 +0000)]
<socket>.getsockopt() and <socket>.setsockopt() can only raise socket.error,
so only catch that specific exception.

24 years agoCatch only the relevant exceptions instead of using a bare except clause.
Fred Drake [Fri, 11 May 2001 18:27:00 +0000 (18:27 +0000)]
Catch only the relevant exceptions instead of using a bare except clause.

24 years agoAdd some text to make the dircmp object section more readable, and move
Fred Drake [Fri, 11 May 2001 17:01:32 +0000 (17:01 +0000)]
Add some text to make the dircmp object section more readable, and move
some stuff around.

24 years agoInclude sys/modem.h if we have it; this is needed on HP-UX to provide
Fred Drake [Fri, 11 May 2001 16:14:17 +0000 (16:14 +0000)]
Include sys/modem.h if we have it; this is needed on HP-UX to provide
constants used by other macros from the headers.

Conditionalize VREPRINT and VDISCARD; these are not available on HP-UX.

This closes bug #417418.

24 years agothe usual...
Fred Drake [Fri, 11 May 2001 16:11:25 +0000 (16:11 +0000)]
the usual...

24 years agoAdd a check for sys/modem.h, needed by termios on HP-UX.
Fred Drake [Fri, 11 May 2001 16:10:56 +0000 (16:10 +0000)]
Add a check for sys/modem.h, needed by termios on HP-UX.

24 years ago--sigh--
Fred Drake [Fri, 11 May 2001 15:49:19 +0000 (15:49 +0000)]
--sigh--
Finish the last set of changes to these files so the conversion does not
break.

24 years agoMarkup adjustments to avoid getting junk in the index.
Fred Drake [Fri, 11 May 2001 15:46:45 +0000 (15:46 +0000)]
Markup adjustments to avoid getting junk in the index.

24 years agoVariant of SF patch 423181
Jeremy Hylton [Fri, 11 May 2001 14:48:41 +0000 (14:48 +0000)]
Variant of SF patch 423181

For rich comparisons, use instance_getattr2() when possible to avoid
the expense of setting an AttributeError.  Also intern the name_op[]
table and use the interned strings rather than creating a new string
and interning it each time through.

24 years agounlink() would normally be found in the "os" module, so use it from there.
Fred Drake [Fri, 11 May 2001 14:29:21 +0000 (14:29 +0000)]
unlink() would normally be found in the "os" module, so use it from there.

Remove unused import of "sys".

If the file TESTFN exists before we start, try to remove it.

Add spaces around the = in some assignments.

24 years agoCosmetic: code under "else" clause was missing indent.
Tim Peters [Fri, 11 May 2001 03:36:45 +0000 (03:36 +0000)]
Cosmetic:  code under "else" clause was missing indent.

24 years agoReplace "\begin{classdesc}{SomeClass}{\unspecified}" with
Fred Drake [Fri, 11 May 2001 01:08:13 +0000 (01:08 +0000)]
Replace "\begin{classdesc}{SomeClass}{\unspecified}" with
"\begin{classdesc*}{SomeClass}" -- the rendering of \unspecified was
identical to \moreargs, so this helps clarify things just a little.

24 years agoDocument the new classdesc* environment, and the previously undocumented
Fred Drake [Fri, 11 May 2001 01:01:12 +0000 (01:01 +0000)]
Document the new classdesc* environment, and the previously undocumented
excclassdesc environment.

24 years agoDefine a new environment, classdesc*, which can be used to document a
Fred Drake [Fri, 11 May 2001 01:00:30 +0000 (01:00 +0000)]
Define a new environment, classdesc*, which can be used to document a
class without providing any information about the constructor.  This
should be used for classes which only exist to act as containers rather
than as factories for instances.

24 years agoWrite a better synopsis for the Scrap module, and provide a link to
Fred Drake [Thu, 10 May 2001 22:37:38 +0000 (22:37 +0000)]
Write a better synopsis for the Scrap module, and provide a link to
useful documentation on the Scrap Manager.

24 years agoActually include a synopsis line for the ColorPicker module.
Fred Drake [Thu, 10 May 2001 22:36:13 +0000 (22:36 +0000)]
Actually include a synopsis line for the ColorPicker module.

24 years agoRestore dicts' tp_compare slot, and change dict_richcompare to say it
Tim Peters [Thu, 10 May 2001 21:45:19 +0000 (21:45 +0000)]
Restore dicts' tp_compare slot, and change dict_richcompare to say it
doesn't know how to do LE, LT, GE, GT.  dict_richcompare can't do the
latter any faster than dict_compare can.  More importantly, for
cmp(dict1, dict2), Python *first* tries rich compares with EQ, LT, and
GT one at a time, even if the tp_compare slot is defined, and
dict_richcompare called dict_compare for the latter two because
it couldn't do them itself.  The result was a lot of wasted calls to
dict_compare.  Now dict_richcompare gives up at once the times Python
calls it with LT and GT from try_rich_to_3way_compare(), and dict_compare
is called only once (when Python gets around to trying the tp_compare
slot).
Continued mystery:  despite that this cut the number of calls to
dict_compare approximately in half in test_mutants.py, the latter still
runs amazingly slowly.  Running under the debugger doesn't show excessive
activity in the dict comparison code anymore, so I'm guessing the culprit
is somewhere else -- but where?  Perhaps in the element (key/value)
comparison code?  We clearly spend a lot of time figuring out how to
compare things.

24 years agoMake test_mutants stronger by also adding random keys during comparisons.
Tim Peters [Thu, 10 May 2001 20:18:30 +0000 (20:18 +0000)]
Make test_mutants stronger by also adding random keys during comparisons.
A Mystery:  test_mutants ran amazingly slowly even before dictobject.c
"got fixed".  I don't have a clue as to why.  dict comparison was and
remains linear-time in the size of the dicts, and test_mutants only tries
100 dict pairs, of size averaging just 50.  So "it should" run in less than
an eyeblink; but it takes at least a second on this 800MHz box.

24 years agoChange test_mmap.py to use test_support.TESTFN instead of hardcoded "foo",
Tim Peters [Thu, 10 May 2001 20:03:04 +0000 (20:03 +0000)]
Change test_mmap.py to use test_support.TESTFN instead of hardcoded "foo",
and wrap the body in try/finally to ensure TESTFN gets cleaned up no
matter what.

24 years agoRepair typos in comments.
Tim Peters [Thu, 10 May 2001 19:40:30 +0000 (19:40 +0000)]
Repair typos in comments.

24 years agoRepair typo in comment.
Tim Peters [Thu, 10 May 2001 18:58:31 +0000 (18:58 +0000)]
Repair typo in comment.

24 years agoChange some text just a little to avoid font-lock hell.
Fred Drake [Thu, 10 May 2001 18:41:02 +0000 (18:41 +0000)]
Change some text just a little to avoid font-lock hell.

24 years agoFix typo in weakref.proxy() documentation.
Fred Drake [Thu, 10 May 2001 17:22:17 +0000 (17:22 +0000)]
Fix typo in weakref.proxy() documentation.
This closes SF bug #423087.

24 years agoExtend the weakref test suite to cover the complete mapping interface for
Fred Drake [Thu, 10 May 2001 17:16:38 +0000 (17:16 +0000)]
Extend the weakref test suite to cover the complete mapping interface for
both weakref.Weak*Dictionary classes.

This closes SF bug #416480.

24 years agoUpdate example to no longer use the FCNTL module.
Fred Drake [Thu, 10 May 2001 15:57:17 +0000 (15:57 +0000)]
Update example to no longer use the FCNTL module.

24 years agoFix the fcntl() docstring so the user is not mis-directed to the FCNTL
Fred Drake [Thu, 10 May 2001 15:54:32 +0000 (15:54 +0000)]
Fix the fcntl() docstring so the user is not mis-directed to the FCNTL
module for useful constants.

24 years agoDo no regenerate modules that should no longer be here.
Fred Drake [Thu, 10 May 2001 15:52:47 +0000 (15:52 +0000)]
Do no regenerate modules that should no longer be here.

24 years agoRemove all remaining uses of the FCNTL module from the standard library.
Fred Drake [Thu, 10 May 2001 15:33:31 +0000 (15:33 +0000)]
Remove all remaining uses of the FCNTL module from the standard library.

24 years agoFix typo reported by David Goodger. This closes SF patch #422383.
Fred Drake [Thu, 10 May 2001 15:09:36 +0000 (15:09 +0000)]
Fix typo reported by David Goodger.  This closes SF patch #422383.

24 years agoRemove all mentions of the strop module -- it has been pronounced Evil.
Fred Drake [Thu, 10 May 2001 15:05:03 +0000 (15:05 +0000)]
Remove all mentions of the strop module -- it has been pronounced Evil.
(The string "strop" is found in the rexec documentation, but that should
not be changed until strop is actually removed or rexec no longer allows
it.)

24 years agoAdded a note that test_longexp needs 400MB.
Jack Jansen [Thu, 10 May 2001 12:20:30 +0000 (12:20 +0000)]
Added a note that test_longexp needs 400MB.

24 years agoHas been dead so long that there's no use keeping it in the active bit of the repository.
Jack Jansen [Thu, 10 May 2001 12:17:03 +0000 (12:17 +0000)]
Has been dead so long that there's no use keeping it in the active bit of the repository.

24 years agoSF bug #422121 Insecurities in dict comparison.
Tim Peters [Thu, 10 May 2001 08:32:44 +0000 (08:32 +0000)]
SF bug #422121 Insecurities in dict comparison.
Fixed a half dozen ways in which general dict comparison could crash
Python (even cause Win98SE to reboot) in the presence of kay and/or
value comparison routines that mutate the dict during dict comparison.
Bugfix candidate.

24 years agoUpdate to reflect deprecation of the FCNTL module: The fcntl module does
Fred Drake [Thu, 10 May 2001 05:17:02 +0000 (05:17 +0000)]
Update to reflect deprecation of the FCNTL module:  The fcntl module does
*not* define O_RDWR; get that from the os module.

24 years agopatch 418489 from Andrew Dalke for string format bug
Steve Purcell [Thu, 10 May 2001 01:28:40 +0000 (01:28 +0000)]
patch 418489 from Andrew Dalke for string format bug

24 years agoGuido has Spoken. Restore strop.replace()'s treatment of a 0 count as
Tim Peters [Thu, 10 May 2001 01:23:39 +0000 (01:23 +0000)]
Guido has Spoken.  Restore strop.replace()'s treatment of a 0 count as
meaning infinity -- but at least warn about it in the code!  I pissed
away a couple hours on this today, and don't wish the same on the next
in line.
Bugfix candidate.

24 years agoThe strop module and test_strop.py believe replace() with a 0 count
Tim Peters [Thu, 10 May 2001 00:59:45 +0000 (00:59 +0000)]
The strop module and test_strop.py believe replace() with a 0 count
means "replace everything".  But the string module, string.replace()
amd test_string.py believe a 0 count means "replace nothing".
"Nothing" wins, strop loses.
Bugfix candidate.

24 years agoHeh. I need a break. After this: stropmodule & stringobject were more
Tim Peters [Thu, 10 May 2001 00:32:57 +0000 (00:32 +0000)]
Heh.  I need a break.  After this:  stropmodule & stringobject were more
out of synch than I realized, and I managed to break replace's "count"
argument when it was 0.  All is well again.  Maybe.
Bugfix candidate.

24 years agoFudge. stropmodule and stringobject both had copies of the buggy
Tim Peters [Thu, 10 May 2001 00:05:33 +0000 (00:05 +0000)]
Fudge.  stropmodule and stringobject both had copies of the buggy
mymemXXX stuff, and they were already out of synch.  Fix the remaining
bugs in both and get them back in synch.
Bugfix release candidate.

24 years agoSF bug #422088: [OSF1 alpha] string.replace().
Tim Peters [Wed, 9 May 2001 23:00:26 +0000 (23:00 +0000)]
SF bug #422088: [OSF1 alpha] string.replace().
Platform blew up on "123".replace("123", "").  Michael Hudson pinned the
blame on platform malloc(0) returning NULL.
This is a candidate for all bugfix releases.

24 years agoMechanical changes for easier edits.
Tim Peters [Wed, 9 May 2001 22:15:03 +0000 (22:15 +0000)]
Mechanical changes for easier edits.

24 years agoRemove the old platform-specific FCNTL.py modules; these are no longer
Fred Drake [Wed, 9 May 2001 21:15:06 +0000 (21:15 +0000)]
Remove the old platform-specific FCNTL.py modules; these are no longer
needed now that fcntl exports the constants.

24 years agoAdd a new FCNTL.py backward compatibility module that issues a deprecation
Fred Drake [Wed, 9 May 2001 21:13:23 +0000 (21:13 +0000)]
Add a new FCNTL.py backward compatibility module that issues a deprecation
warning.  This is similar to the TERMIOS backward compatbility module.

24 years agoUpdate the tests for the fcntl module to check passing in file objects,
Fred Drake [Wed, 9 May 2001 21:11:59 +0000 (21:11 +0000)]
Update the tests for the fcntl module to check passing in file objects,
and using the constants defined there instead of FCNTL.

24 years agoUpdate the fcntl module documentation.
Fred Drake [Wed, 9 May 2001 21:09:57 +0000 (21:09 +0000)]
Update the fcntl module documentation.

24 years agoModify to allow file objects wherever file descriptors are needed.
Fred Drake [Wed, 9 May 2001 21:02:02 +0000 (21:02 +0000)]
Modify to allow file objects wherever file descriptors are needed.
This closes SF bug #231328.

Added all constants needed to use the functions defined in this module
that are not defined elsewhere (the O_* symbols are available in the
os module).  No additonal modules are needed to use this now.

24 years agofdconv(): Do not second guess the error condition returned by
Fred Drake [Wed, 9 May 2001 20:14:09 +0000 (20:14 +0000)]
fdconv():  Do not second guess the error condition returned by
    PyObject_AsFileDescriptor() -- it does the same thing everywhere, so
    use it the same way everyone else does so that exceptions are
    consistent.  This means we have less code here, and we do not need to
    resort to hackish ways of getting the Python-visible function name to
    fdconv().

24 years agoSF patch #416247 2.1c1 stringobject: unused vrbl cleanup.
Tim Peters [Wed, 9 May 2001 20:06:00 +0000 (20:06 +0000)]
SF patch #416247 2.1c1 stringobject: unused vrbl cleanup.
Thanks to Mark Favas.

24 years agoItamar Shtull-Trauring <python@itamarst.org>:
Fred Drake [Wed, 9 May 2001 19:57:37 +0000 (19:57 +0000)]
Itamar Shtull-Trauring <python@itamarst.org>:
Updates zipfile.ZipFile docs to mention the fact that you can create a
ZipFile instance from an arbitrary file-like object.

This closes patch #418011.

24 years agoThree uses of makesockaddr() used sockaddr buffers that had not be cleared;
Fred Drake [Wed, 9 May 2001 19:11:33 +0000 (19:11 +0000)]
Three uses of makesockaddr() used sockaddr buffers that had not be cleared;
this could cause invalid paths to be returned for AF_UNIX sockets on some
platforms (including FreeBSD 4.2-RELEASE), appearantly because there is
no assurance that the address will be nul-terminated when filled in by the
kernel.

PySocketSock_recvfrom():  Use PyString_AS_STRING() to get the data pointer
    of a string we create ourselves; there is no need for the extra type
    check from PyString_AsString().

This closes SF bug #416573.

24 years agoSF patch #416249, from Mark Favas: 2.1c1 compile: unused vrbl cleanup
Tim Peters [Wed, 9 May 2001 18:53:51 +0000 (18:53 +0000)]
SF patch #416249, from Mark Favas:  2.1c1 compile: unused vrbl cleanup

24 years agoMinor fiddling related to
Tim Peters [Wed, 9 May 2001 18:48:26 +0000 (18:48 +0000)]
Minor fiddling related to
SF patch 416251 2.1c1 mmapmodule: unused vrbl cleanup

24 years agoUpdate build notes for Mac OS X 10.0.
Fred Drake [Wed, 9 May 2001 18:13:47 +0000 (18:13 +0000)]
Update build notes for Mac OS X 10.0.
This closes SF bug #416530.

24 years agoOnly import termio.h on OSF, and add a comment about why it is needed there.
Fred Drake [Wed, 9 May 2001 17:53:06 +0000 (17:53 +0000)]
Only import termio.h on OSF, and add a comment about why it is needed there.
This header does not exist on all Unix flavors; FreeBSD in particular does
not include it.

This closes SF bug #422320.

24 years agoFix the operator precedence table: exponentiation binds tighter than
Fred Drake [Wed, 9 May 2001 16:51:49 +0000 (16:51 +0000)]
Fix the operator precedence table: exponentiation binds tighter than
negation.

This closes SF bug #421999.

24 years agoRemove items that have been done or are being tracked in the SourceForge
Fred Drake [Wed, 9 May 2001 16:43:47 +0000 (16:43 +0000)]
Remove items that have been done or are being tracked in the SourceForge
bug tracker.

24 years agoUpdate the directory names to match changes at SourceForge.
Fred Drake [Wed, 9 May 2001 16:33:34 +0000 (16:33 +0000)]
Update the directory names to match changes at SourceForge.

24 years agoTrivial tests of urllib2 for recent SF bug
Jeremy Hylton [Wed, 9 May 2001 15:50:25 +0000 (15:50 +0000)]
Trivial tests of urllib2 for recent SF bug

24 years agoWork around limitations of the module synopsis table generation to avoid
Fred Drake [Wed, 9 May 2001 15:50:17 +0000 (15:50 +0000)]
Work around limitations of the module synopsis table generation to avoid
leaking LaTeX2HTML's internal string munging.

This fixes SF bug #420399.

24 years agoRaise useful exception when called with URL for which request type
Jeremy Hylton [Wed, 9 May 2001 15:49:24 +0000 (15:49 +0000)]
Raise useful exception when called with URL for which request type
cannot be determined.

Pseudo-fix for SF bug #420724

24 years agoMinor adjustments to HTML for the module synopsis tables.
Fred Drake [Wed, 9 May 2001 15:32:14 +0000 (15:32 +0000)]
Minor adjustments to HTML for the module synopsis tables.

24 years agoSheesh -- repair the dodge around "cast isn't an lvalue" complaints to
Tim Peters [Wed, 9 May 2001 08:43:21 +0000 (08:43 +0000)]
Sheesh -- repair the dodge around "cast isn't an lvalue" complaints to
restore correct semantics.

24 years agoMark Favas reported that gcc caught me using casts as lvalues. Dodge it.
Tim Peters [Wed, 9 May 2001 07:37:07 +0000 (07:37 +0000)]
Mark Favas reported that gcc caught me using casts as lvalues.  Dodge it.

24 years agoJob.build_html(): Be more robust in ensuring about.html exists; copying
Fred Drake [Wed, 9 May 2001 04:03:16 +0000 (04:03 +0000)]
Job.build_html():  Be more robust in ensuring about.html exists; copying
    the right HTML file to the name about.html is needed even if the
    --numeric option was not given -- some other name may have been
    assigned due to some non-determinism in the algorithm use to perform
    name allocation.  ;-(

This closes the "About..." portion of SF bug #420216.

24 years agoThere is no IMAP class in the imaplib module; the class is IMAP4.
Fred Drake [Wed, 9 May 2001 03:49:48 +0000 (03:49 +0000)]
There is no IMAP class in the imaplib module; the class is IMAP4.
There is no imap module; refer to imaplib instead, since it exists.
Move the "See Also:" section in front of the sub-sections, for
consistency with other portions of the library reference.

This closes the library reference portion of SF bug #420216.

24 years agoNote that when inplace=1 existing backup files will be removed silently.
Fred Drake [Wed, 9 May 2001 03:24:55 +0000 (03:24 +0000)]
Note that when inplace=1 existing backup files will be removed silently.
Closes SF bug #420230.

24 years agoAlways pass a full path name to LoadLibraryEx(). Fixes some Windows 9x problems...
Mark Hammond [Wed, 9 May 2001 00:50:59 +0000 (00:50 +0000)]
Always pass a full path name to LoadLibraryEx().  Fixes some Windows 9x problems.  As discussed on python-dev

24 years agoAck! Restore the COUNT_ALLOCS one_strings code.
Tim Peters [Wed, 9 May 2001 00:31:40 +0000 (00:31 +0000)]
Ack!  Restore the COUNT_ALLOCS one_strings code.

24 years agoMy change to string_item() left an extra reference to each 1-character
Tim Peters [Wed, 9 May 2001 00:24:55 +0000 (00:24 +0000)]
My change to string_item() left an extra reference to each 1-character
interned string created by "string"[i].  Since they're immortal anyway,
this was hard to notice, but it was still wrong <wink>.

24 years agoIntern 1-character strings as soon as they're created. As-is, they aren't
Tim Peters [Tue, 8 May 2001 22:33:50 +0000 (22:33 +0000)]
Intern 1-character strings as soon as they're created.  As-is, they aren't
interned when created, so the cached versions generally aren't ever
interned.  With the patch, the
Py_INCREF(t);
*p = t;
Py_DECREF(s);
return;
indirection block in PyString_InternInPlace() is never executed during a
full run of the test suite, but was executed very many times before.  So
I'm trading more work when creating one-character strings for doing less
work later.  Note that the "more work" here can happen at most 256 times
per program run, so it's trivial.  The same reasoning accounts for the
patch's simplification of string_item (the new version can call
PyString_FromStringAndSize() no more than 256 times per run, so there's
no point to inlining that stuff -- if we were serious about saving time
here, we'd pre-initialize the characters vector so that no runtime testing
at all was needed!).