[Fwd: Re: [Merge] lp:~jameinel/bzr/2.1.0b3-win32-shell-completion into lp:bzr]

[Alexander Belchenko]

John Arbash Meinel пишет:
> Alexander Belchenko wrote:
>> -------- Исходное сообщение --------
>> Тема: Re: [Merge] lp:~jameinel/bzr/2.1.0b3-win32-shell-completion into	lp:bzr
>> Дата: Thu, 05 Nov 2009 20:49:14 -0000
>> От: Alexander Belchenko <bialix at ukr.net>
>> Отвечать: mp+14445 at code.launchpad.net
>> Кому: John A Meinel <john at arbash-meinel.com>
>>
>> Review: Needs Fixing
>> I've late here but I want to say my word.
>>
>> John, I agree with Martin that using single quotes is bad idea. I hope you decided to remove them.
>> All Windows people used to double quotes so single quotes is not Windows way.
>>
>> Also I think will be nice to have a way to disable this glob expansion via environment variable,
>> e.g. BZR_GLOB=0 or BZR_GLOB=no. Using configs there is wrong idea, but os.environ should be OK and fast.
> 
> I certainly think it is possible. Are you sure it is *desirable* ?

I think with your patch there is no easy way to compare windows shell 
behavior with bzr command-line parsing behavior for real-world data. 
Specifically I'm still trying to figure out how it will affects 
qsubprocess in QBzr: we don't specially quote bencoded stream now (I 
think Qt should do this if needed), maybe it's our fault, but looking at 
your tests it seems we can get sometimes wrong behavior with qcommit and 
some commit messages with mix of single and double quotes. Maybe I'm 
just plain wrong.

So if there will be BZR_GLOB knob to switch between old and new way it 
will be much simpler to debug corner cases if they will appear suddenly 
on Windows. Do you think it's desirable?