Barry Alan Scott added the comment:
Thanks Petr, I'll watch for the PEP.
FYI: I work on the assumption that if I use Py_LIMITED_API and the header
files provide an API guarded by an #if then its "offical".
--
___
Python track
Barry Alan Scott added the comment:
Is there a test suite that checks that the limited API functions can all be
linked against?
--
___
Python tracker
<https://bugs.python.org/issue43
Barry Alan Scott added the comment:
Understood that changing the status quo may have problems.
Maybe you do an HTTP redirect from the old name to the new name?
Or have the installed available under both names but only advertise the new
name
Barry Alan Scott added the comment:
I was wondering what the bottom text box was for.
The usual UX for check boxes is to toggle when the text of the control is
clicked.
Because of that I avoid clicking on check boxes that I do not know if I want to
change. Hence I did not get to see
Barry Alan Scott added the comment:
The readme page is a good 2nd best is PKG install has not option to control the
text on the INstallation type page.
My guess is that the info would need be in visible without
scrolling for people to notice it. (UX is hard
Barry Alan Scott added the comment:
>From "UNIX command-line tools" its not clear to me that that prevents
symlinks in /usr/local/bin. Maybe word it as "Add python to /usr/local/bin".
Is this documented anywhere?
I just looked at the download page https://www.pyt
Barry Alan Scott added the comment:
It was not obvious that I could do that with Customise.
Indeed I installed 3.10a5 just so that I could remind myself of what the
installer did.
It would help is the Installation type page told the user why they might want
to use the Customise
New submission from Barry Alan Scott :
The popular homebrew system installs its binaries into /usr/local/bin.
The macOS python install also defaults to /usr/local/bin where it
puts symlinks to Python like:
% ls -l /usr/local/bin/python3.10
lrwxr-xr-x 1 root 73 8 Feb 19:45:50 2021 /usr
New submission from Barry Alan Scott :
By default the python macOS installer will edit .bash_profile and .zsh_profile
to add Python to the path.
For many users this is a problem that requires the added lines to be
removed after each install.
I suggest that there is a mechanism that the user
Barry Alan Scott added the comment:
I happy with your names. I think they are better then my version.
--
___
Python tracker
<https://bugs.python.org/issue43
New submission from Barry Alan Scott :
New users of Python on Windows frequently confuse the Python installation
program with Python itself.
They also can end up with the wrong 32 or 64 bit version.
I suggest that the installation program name is changed to include "setup" in
New submission from Barry Alan Scott :
When using the limited API PyCMethod_New is not defined in python3.lib.
This leads to a link error.
This error is in 3.9 and 3.10a5
Discovered when I improved the PyCXX test suite.
Example output:
Link obj\simple.pyd
cl /nologo /W4 /LD /Zi /MT
Barry Alan Scott added the comment:
Ok. I'll add this to my todo list.
--
status: pending -> open
___
Python tracker
<https://bugs.python.org/issue1
Barry Alan Scott added the comment:
Anthony,
Now that everything is opened using open_code that returns bytes its
not clear to me why this breaks for you.
Further the data must be bytes for the codings to be figured out.
Removing the b'\n' may be reasonable, but not for the reason given.
I
Barry Alan Scott added the comment:
I have pushed the fix onto https://github.com/python/cpython/pull/19595
with an API test case and the changes to keep the debain subclassing
working.
I'm new the the work flow. Let me know if I need to change anything
Change by Barry Alan Scott :
--
pull_requests: +18931
pull_request: https://github.com/python/cpython/pull/19595
___
Python tracker
<https://bugs.python.org/issue40
Barry Alan Scott added the comment:
I have the fix coded and tested.
I run out of git knowledge to update the PR branch so that I can push the fix.
I'll work on it more later in the day.
--
___
Python tracker
<https://bugs.python.org/issue40
Change by Barry Alan Scott :
--
pull_requests: +18895
stage: needs patch -> patch review
pull_request: https://github.com/python/cpython/pull/19549
___
Python tracker
<https://bugs.python.org/issu
Barry Alan Scott added the comment:
Regarding test case. I will need to know what pymindeps is doing to be able to
design a suitable test case.
--
___
Python tracker
<https://bugs.python.org/issue40
Barry Alan Scott added the comment:
I need to see the code of pymindeps to understand what you are doing and how to
fix this. Can you post a URL to the source please?
Are you aware that load_module() changed in other ways that are required to fix
the bug?
You may have to change yout
Barry Alan Scott added the comment:
io.open_code() used in the patch.
--
components: -Windows
___
Python tracker
<https://bugs.python.org/issue40260>
___
___
New submission from Barry Alan Scott :
modulefinder.py does not open source files in "rb" which
prevents compile() from applying the encoding rules.
This first showed up for me on Windows with Python 3.8.
Here is my test case and the results on Windows with 3.8.
import mo
Barry Alan Scott added the comment:
I should point out using Python2.7 with wxPython I do not see this issue.
This issue was exposed during my efforts to port pysvn from py2.7+wxPython to
py3.4+wxpython-phoenix.
I do setup locale on the main thread very early before starting the background
Barry Alan Scott added the comment:
Mac OS X use the __CF_USER_TEXT_ENCODING env var to setup the locale in for
native libraries.
I found that for GUI python code I needed to convert the value in
__CF_USER_TEXT_ENCODING into a suitable call to setlocale().
The code I use is attached
New submission from Barry Alan Scott:
I'm seeing a random traceback when starting a new thread on Mac OS X 10.10.2
with python 3.4.3 final.
Exception in thread Thread-1:
Traceback (most recent call last):
File
/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/threading.py
Barry Alan Scott added the comment:
Why not use IDLE? Workbench is a lot of code and dependencies.
I expect that it works because idle.app was created using the --no-zipimport
option that is new in 2.7.
However with zip import the code is badly broken.
Build IDLE.app with zip import and you
New submission from Barry Alan Scott:
bundlebuild in pytthon 2.7 fails to create usable applications.
After examining the code there are a number of coding errors:
1. optimize will puts the -O in the wrong place in the argv
2. Modules.zip is never added to the path
The bunderbuilder in 2.6
Barry Alan Scott added the comment:
I'm using the pysvn workbench sources to find the problem.
Test cases for Mac Apps are not that simple. You need a GUI
framework for starters.
But I'm happy to test any candidate fix using workbench.
As I said a code inspection of the changes since 2.6
Barry Alan Scott barry-sc...@users.sourceforge.net added the comment:
I'm happy to review patches or create them for you.
I see a related set of errors on Mac OS X that are down to issues in
bytes_methods.py that mirror the pyport.h issues.
In C++ isspace and friends are functions not macros
Barry Alan Scott barry-sc...@users.sourceforge.net added the comment:
I have wxPython installed. Uninstalling it changes the order of sys.path to
this:
['',
'/Library/Frameworks/Python.framework/Versions/2.7/lib/python27.zip',
'/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7
New submission from Barry Alan Scott barry-sc...@users.sourceforge.net:
I'm testing on Mac OS X 10.7.3
http://bugs.python.org/issue4865 added
/Library/Python/2.7/site-packages
to the path for python.org built 2.7.
This will work for .py files but .so cannot be shared.
pysvn will SEGV
Barry Alan Scott barry-sc...@users.sourceforge.net added the comment:
Pysvn binary kits can be downloaded from
http://pysvn.tigris.org/files/documents/1233/49180/py27_python_org_pysvn_svn1617-1.7.6-1457-intel.dmg
and
http://pysvn.tigris.org/files/documents/1233/49178
Barry Alan Scott barry-sc...@users.sourceforge.net added the comment:
Updated patch with comment explaining cast.
--
Added file: http://bugs.python.org/file15184/const_api_r75619.patch
___
Python tracker rep...@bugs.python.org
http://bugs.python.org
Barry Alan Scott barry-sc...@users.sourceforge.net added the comment:
I was trying to avoid changing the const ness of output parameters.
Given the advice not to go mad on putting const everywhere.
1) I can comment the casts to maintain this goal.
2) Or change the output paramter const ness
Barry Alan Scott barry-sc...@users.sourceforge.net added the comment:
Here is my 1st patch to allow const char * and const Py_UNICODE *
It is agsinst http://svn.python.org/projects/python/trunk.
What I have attempted to do is stop the public API of python needing
char * or Py_UNICODE * where
Barry Alan Scott barry-sc...@users.sourceforge.net added the comment:
I'd guess that this change can first be made against 2.7 and 3.2
so that API's do not change.
Where do I find the source code to generate the patch against
for 2.7 and 3.2
New submission from Barry Alan Scott barry-sc...@users.sourceforge.net:
Many Python API functions are causing GCC to rightly complain
that const char * strings are being passed to python functions
that take char * but do not need to modify the args.
g++ reports example.cxx:633: warning
New submission from Barry Alan Scott barry-sc...@users.sourceforge.net:
On Mac OS X 10.5
$ LC_ALL=ru_RU.koi8-r python3.0 -c 'import time;print( time.strftime(%A))'
Traceback (most recent call last):
File string, line 1, in module
UnicodeDecodeError: 'utf8' codec can't decode bytes in position
New submission from Barry Alan Scott barry-sc...@users.sourceforge.net:
The result of time.strftime seems to be in the locale encoding
but this is not pointed out in the documentation.
Ideally an example like this would be in the docs to show how to
deal with strftime output:
import locale
Barry Alan Scott [EMAIL PROTECTED] added the comment:
You are right that its the Py::Long( Py::Float( double( x ) ) )
that is triggering this problem.
Here is the gdb from the powerpc build that shows the
info you asked for and show res being corrupt.
I'm going to try and build a smaller
Barry Alan Scott [EMAIL PROTECTED] added the comment:
O.k. I know what is going on.
Here is the description from abstracts.h for PyNumber_Long:
PyAPI_FUNC(PyObject *) PyNumber_Long(PyObject *o);
/*
Returns the o converted to a long integer object on success
Barry Alan Scott [EMAIL PROTECTED] added the comment:
My concern and yours is that this is not backwards
compatible. I would hate to see random failures
of extensions written using PyCXX because of this.
I'm tempted to says that I'll keep PyCXX 5.x as is for
Python 2.x and leave all the changes
New submission from Barry Alan Scott [EMAIL PROTECTED]:
I am testing PySVN against python2.6b3.
I see a failure when PyNumber_Long is called with a Float.
It raises TypeError.
The same code works on 2.3, 2.4 and 2.5.
Looking with GDB I see:
(gdb) bt
#0 PyNumber_Long (o=0x1809384) at Objects
Barry Alan Scott [EMAIL PROTECTED] added the comment:
I have Xcode 2.3 which is quite old.
Simple program to test assert works with my current setup.
I'm going to update to Xcode 2.5 and see what happens.
I'll report back once I've installed and rebuild python.
Added file: http
Barry Alan Scott [EMAIL PROTECTED] added the comment:
Xcode 2.5 solves the build issue.
I suspect you can close the 1099 saying use Xcode 2.5.
I think you can only get 2.5 by login in to the Apple developer site.
___
Python tracker [EMAIL PROTECTED]
http
Barry Alan Scott [EMAIL PROTECTED] added the comment:
Grr... this problem is a pain... I have the __eprint undefined back...
I tried to get readline going as well and rebuilt. Couldn't get that
working then rebuilt without readline stuff and got the __eprintf.
I'm goint to have to give up
New submission from Barry Alan Scott [EMAIL PROTECTED]:
I wanted to use Py_DEBUG build to help debug a problem
with ref counts in a C++ extension.
I cannot find eprintf in the sources of python
where does this symbol come from? How do I fix the
build to define it?
$ sw_vers
ProductName
Barry Alan Scott [EMAIL PROTECTED] added the comment:
As far as I know I'm using the Xcode compiler. Does this match
your expectations?
$ which gcc
/usr/bin/gcc
$ gcc -v
Using built-in specs.
Target: powerpc-apple-darwin8
Configured with: /private/var/tmp/gcc/gcc-5341.obj~1/src/configure
New submission from Barry Alan Scott [EMAIL PROTECTED]:
$ sw_vers
ProductName:Mac OS X
ProductVersion: 10.4.11
BuildVersion: 8S165
$ python3.0
Python 3.0b2 (r30b2:65080, Jul 20 2008, 08:46:13)
[GCC 4.0.1 (Apple Computer, Inc. build 5341)] on darwin
Type help, copyright, credits
Barry Alan Scott [EMAIL PROTECTED] added the comment:
Fixing 2to3 with the full path to the fixes folder gets this traceback:
$ ./2to3 /dev/null
Traceback (most recent call last):
File ./2to3, line 5, in module
sys.exit(refactor.main(/Library/Frameworks/Python.framework//Versions/3.0/lib
New submission from Barry Alan Scott [EMAIL PROTECTED]:
$ sw_vers
ProductName:Mac OS X
ProductVersion: 10.4.11
BuildVersion: 8S165
./configure --enable-framework
make
...
make install
...
DYLD_FRAMEWORK_PATH=/Users/barry/Work/Python-3.0b1: ../python.exe
./scripts/BuildApplet.py
Barry Alan Scott [EMAIL PROTECTED] added the comment:
I don't see beta 2 on python.org or I'd have used it...
___
Python tracker [EMAIL PROTECTED]
http://bugs.python.org/issue3398
52 matches
Mail list logo