Python shell wont open idle or an exisiting py file

Terry Reedy tjreedy at udel.edu
Sat Feb 1 03:51:48 EST 2014


On 2/1/2014 2:26 AM, Chris Angelico wrote:
> On Sat, Feb 1, 2014 at 4:46 PM, Terry Reedy <tjreedy at udel.edu> wrote:
>> On 1/31/2014 10:36 PM, Chris Angelico wrote:
>>>
>>> On Sat, Feb 1, 2014 at 1:54 PM, MRAB <python at mrabarnett.plus.com> wrote:
>>>>
>>>> I think that some years ago I heard about a variation on UTF-8
>>>> (Microsoft?) where codepoint U+0000 is encoded as 0xC0 0x80 so that the
>>>> null byte can be used as the string terminator.
>>>>
>>>> I had a look on Wikipedia found this:
>>>>
>>>> http://en.wikipedia.org/wiki/Null-terminated_string
>>>
>>>
>>> Yeah, it's a common abuse of UTF-8. It's a violation of spec, but an
>>> understandable one. However, I don't understand why the first part -
>>> why should \0 become U+0000 but (presumably) the \a later on
>>> (...cs\accel...) doesn't become U+0007, etc?
>>
>>
>> Because only  \0 has a special meaning in a C string,

I should have added 'to C itself', as the string terminator.

>> and Tk is written in C and uses C strings.
>
> Eh? I've used \a in C programs (not often but I have used it).
>
> It's possible that \0 is the only one that actually bombs anything
> (because of C0 80 representation).

\0 can bomb C byte processing by terminating it sooner than it should. 
Its unexpected replacement bombs utf-8 decoding.

 > But since \7 and \a both represent
> 0x07 in a C string, I would expect there to be other problems, if it's
> interpreting it as source. Ah well! Weird weird.

While other control codes may have special meaning to a terminal or 
other device, to do not have special meaning to the operation of C 
string functions themselves (except possible for a 'getline' function 
looking for n -- but I do not remember is the C stdlib has any such 
functions).

I am speaking from my memory of C. I have not looked at the Tk C code to 
see just what it did where to create the exception. I am just happy that 
Serhiy was able to fixed tkinter without causing another test to fail.

-- 
Terry Jan Reedy




More information about the Python-list mailing list