| msg12526 - (view) |
Author: Brian Lenihan (brianl) |
Date: 2002-09-29 00:20 |
| My OS X laptop did not have LANG set in the bash envrionment and idle (AquaTK) failed to launch with an AttributeError in IOBinding.py try: encoding = locale.getdefaultlocale()[1] codecs.lookup(encoding) except (ValueError, LookupError): pass encoding = encoding.lower() ^^^^^ encoding is None here because locale.getdefaultlocale() returns (None, None) when the locale can't be determined. After finding that adding LANG=en_US to my .bashrc didn't help, I Iearned OS X GUI apps don't use the shell's environment, but do use whatever they find in ~/.MacOSX/environment.plist. Adding a couple lines: LANG en_US makes idle happy. An alternative would be to make sure encoding never gets set to None. --- IOBinding.py.orig Sat Sep 28 17:17:02 2002 +++ IOBinding.py Sat Sep 28 17:17:42 2002 @@ -64,7 +64,7 @@ # which may give a clue. Unfortunately, getdefaultlocale has # bugs that can cause ValueError. try: - encoding = locale.getdefaultlocale()[1] + encoding = locale.getdefaultlocale()[1] or 'ascii' codecs.lookup(encoding) except (ValueError, LookupError): pass |
|
|
| msg12527 - (view) |
Author: Martin v. Löwis (loewis) *  |
Date: 2002-09-29 22:17 |
| Logged In: YES user_id=21627 Can you please find out how to determine the user's prefererred encoding on OS X? Yes, getdefaultlocale is broken, but that is by design, and cannot be fixed. So we should reduce its usage instead of trying to correct it. |
|
|
| msg12528 - (view) |
Author: Brian Lenihan (brianl) |
Date: 2002-09-30 02:18 |
| Logged In: YES user_id=17671 I've done a little digging and have yet to find a satisfactory answer. This article: http://developer.apple.com/internet/macosx/perl.html has instructions for upgrading perl to 5.8.0 "OS X has one of the necessary environment variables set (LANG), but without the above, you'll get annoying error messages each time you try to run a script. The above assumes the tcsh shell; if you prefer bash, use echo "export LC_ALL=C" >> ~/.bash_profile instead." I'll keep digging for a good answer, but I still believe there is a bug here. If you look in IOBinding.py, you'll see that encoding is given a default value of "ascii". It should leave the section with that value intact or with ascii replaced with the proper locale. The possibility that getdefaultlocale will punt and return (None, None) is not taken into account in IOBinding.py |
|
|
| msg12529 - (view) |
Author: Martin v. Löwis (loewis) * |
Date: 2002-09-30 07:38 |
| Logged In: YES user_id=21627 I agree that there is a bug somewhere. I also suspect that the bug is in IDLE, but I think it is wrong to fall-back to ASCII on OS X (I'm almost certain that the user's encoding is never just ASCII on that system). I suspect that it is MacRoman or some such, but I also suspect that this either might vary with the localized version of OS X (assuming there are localized versions), or with a per-user setting. |
|
|
| msg12530 - (view) |
Author: Brian Lenihan (brianl) |
Date: 2002-09-30 08:35 |
| Logged In: YES user_id=17671 I was surprised falling back to ascii actually worked. I found a bunch of docs at the devoper.apple.com site. As far as I can tell, the proper way to fix this is to add the necessary support to macglue.h, etc. so the functions in _locale work on OS X as well. I found out why this only just now bit me: I had compatible libs in /usr/local/lib which made nl_langinfo() available in _locale until I upgraded to 10.2.1. The proper fix is going to require some effort by someone who understands Apple API's much better than I do. Since falling back to ascii works (at least for en_US) as you intended - you were the one who added the locale code to IOBinding.py - perhaps a temporary fix is to make sure the fallback works as you intended? |
|
|
| msg12531 - (view) |
Author: Martin v. Löwis (loewis) * |
Date: 2002-09-30 08:46 |
| Logged In: YES user_id=21627 Feel free to use any work-around you feel comfortable with. However, I won't approve a fix that I know is wrong. |
|
|
| msg12532 - (view) |
Author: Brian Lenihan (brianl) |
Date: 2002-10-04 06:36 |
| Logged In: YES user_id=17671 This may as well be closed. I see in the latest Python-dev summary that Guido asked for all future changes to be done in the idle-fork project. Tony Lownds checked in a work-around to the idle-fork project on 09/23 which is similar to what I was suggesting. |
|
|
| msg12533 - (view) |
Author: Martin v. Löwis (loewis) * |
Date: 2002-10-04 07:28 |
| Logged In: YES user_id=21627 You can probably guess that I dislike the change being done to IDLE. Oh well. |
|
|
| msg12534 - (view) |
Author: Brian Lenihan (brianl) |
Date: 2002-10-04 07:53 |
| Logged In: YES user_id=17671 I apologize it this comment reopens this bug. Inititially, I had a hard time understanding your objection, given that my suggestion would have merely made the code behave as you had intended when you wrote it. Upon reflection, however, I understand why you are objecting. |
|
|
| msg12535 - (view) |
Author: Martin v. Löwis (loewis) * |
Date: 2002-10-04 09:21 |
| Logged In: YES user_id=21627 I guess we can leave this closed until a MacOS expert comes along... To reiterate my opposition: I think it is wrong to use 'ascii' as the IDLE encoding on OS X. Even though OS X does support ascii, it supports more than that. As a result, people typing non-ASCII characters will find that IDLE cannot process them properly. This is unfortunate, as IDLE poses a restriction which no foundation in the system. |
|
|