[rfc][patch] bzr whoami can set your branch identity

[Robey Pointer]

On 8 Jan 2006, at 20:18, John Arbash Meinel wrote:

> Robey Pointer wrote:
>>
>> * Unless I misunderstood how it works, I think 'whoami' output should
>> be encoded into the user's encoding, so I fixed that.
>
> This might not be bzrlib.user_encoding. It should take
> sys.stdout.encoding if it is present. (Again when I finish my encoding
> branch, this will be done, I'm through about 1/2 of the commands).

Yeah, sys.stdout.encoding is probably better.


>> * I made a decorator '@with_user_encoding' for running a test  
>> with  the
>> user's encoding temporarily changed to something else.  (I wanted  to
>> test that both ascii and utf-8 would work.)  Is there a better  
>> way  to
>> do that?  I feel like I probably just overlooked some existing
>> mechanism for testing with a temporary encoding.
>
> Again, in my encoding branch, I have a parameter that you can pass to
> run_bzr which sets the encoding.
> Oh, and I generally avoid the Blackbox version of 'runbzr'.  
> bzrlib.tests
> has a TestCase which has 'run_bzr' which I think has a much nicer
> interface. 'runbzr' is just a hangover of when we only had a single
> 'testbzr' file.

It looks like runbzr is just a very thin wrapper around run_bzr so  
it's all good (though the names did cause me some confusion at first).

Once we're able to set stdin/stdout encoding separately, the run_bzr  
interface should accept an option for that too.  The "whoami" tests  
are actually testing different stdout encodings.


> Just to point you to my branch:
> http://bzr.arbash-meinel.com/branches/bzr/encoding/

Cool, I'm downloading it now, though it's gonna be a while before it  
finishes.  If it's straightforward, I'll make my patch against that  
branch instead.


> By the way, Apple Mail doesn't seem to do a very good job with
> attachments. It attaches your patch as:
> application/octet-stream; x-unix-mode=0644; name="whoami.patch"

I think this is because it's trying to guess the file type via  
extension, since I created the file from a shell.  I'll try a .txt  
extension next time.


> Anyway, to discuss you patch...
> I would like a way to specify that I want to set the global option. In
> fact, I might prefer to have it set the global option, unless I use a
> flag to say I want the local one.
> For people just getting started, who just had bzr not correctly set
> their email. They may have started a branch, and found out that the
> email was wrong. So they go to set their email and everything seems
> fine, but then when they switch to another branch, their email is
> incorrect again.

The reason I did it only as a branch option is that the global option  
is stored in bazaar.conf, which I was told shouldn't be changed by  
bzr, only read.  Otherwise I agree with you that it really should be  
setting the global identity.

How do people feel about relaxing that restriction on bazaar.conf, at  
least for the "whoami" command?


> +        self.log('run bzr: %s', ' '.join(argv).encode('string- 
> escape'))
>
> Just do:
> self.log('run bzr: %r', ' '.join(argv))
>
> %r says to use python's repr() function (representation), which for
> strings means it makes them ascii safe.
> Basically %r (repr) is meant something like "give me the form that I
> would put inside a .py file if I wanted to create this".

Yeah, I know. :)  My assumption is generally that the original author  
wrote the code they wanted, so I try to keep the spirit of it.   
Adding encode('string-escape') was my way of trying to make as  
minimal change there as possible.  %r is a bit heavier (it adds  
quotes) but I honestly don't care either way -- I would prefer not to  
have to change that code at all. ;)  So hopefully this patch made  
against your encodings branch will be smaller.

robey