[patch][merge] sftp/http tweaks

[Robey Pointer]

On 10 Jan 2006, at 12:53, John Arbash Meinel wrote:

> Robey Pointer wrote:
>>
>> On 9 Jan 2006, at 10:56, John Arbash Meinel wrote:
>>
>>> Robey Pointer wrote:
>>>
>>>> Since my bzr.robey branch only had 2 un-integrated patches in  
>>>> it   (when
>>>> compared to bzr.dev), I blew it away, did a fresh branch from    
>>>> bzr.dev,
>>>> and applied just those 2 patches.
>>>>
>>>> 1516: change http url parsing to use urlparse, and use the  
>>>> ui_factory
>>>> to ask for a password if necessary
>>>> 1517: allow forcing the use of paramiko via environ var; use  
>>>> prefetch
>>>> on paramiko >= 1.5.2"
>>>>
>>>> So I'd like to resubmit those for merge and/or feedback.
>>>>
>>>> The branch is at http://www.lag.net/~robey/bzr.robey/
>>>>
>>>> robey
>>>
>>>
>>> Have you always had dots over the 'o' in your name?
>>>
>>> I just saw that your latest commits were from 'Röbey Pointer'.   
>>> Grepping
>>> through the revision store, I see about 41 commits, and only the   
>>> last 2
>>> with ö. And your email name doesn't have them either.
>>
>>
>> Oops, that was just from testing the 'whoami' stuff earlier. :)  I
>> sometimes use an umlauted-o for the same reason Spinal Tap do it, but
>> it's not actually a part of my name.  No harm either way.
>>
>>
>>> Is there a reason that you check os.environ at the time of  
>>> reading the
>>> file, rather than doing that as part of _get_ssh_vendor()?
>>
>>
>> Not particularly. :)  Your suggestions sound good to me so I applied
>> them: BZR_SSH is checked instead, and can be used to override the
>> vendor check completely.  The paramiko version check is done at the
>> top, and I used getattr instead of hasattr.
>>
>> This change is now @1520 on bzr.robey.
>>
>> robey
>
> Well, I kind of like these changes, but they are mixed in with the
> lsprof changes. So we kind of have to wait for both of them to get +1.
> (I don't really feel like cherrypicking).
> Also, I'm not sure about the revno 1518:
>   a missing final revision spec (aka '1501..')
>   is implicitly a '-1', not a '0'
> I know for 'bzr diff' we went for an explicit None means working tree.
> Which I think is broken in your branch (since it would get the -1,
> before getting a chance to detect that it was None).
>
> So I guess I would +1 revisions 1516, 1517, and 1520. If you reverted
> 1518, and get +2 to the --lsprof changes, I'll merge your branch.
>
> (To revert a single revision, you can do: bzr merge -r 1518..1517 .,
> this will undo the changes, and then you can 'bzr commit -m "reverted
> 1518"')

The "diff" patch is superceded by the later patch (which I think you  
already have in jam-integration now), and I have the lsprof changes  
elsewhere, so I'll just uncommit back to 1517 and apply the new patch  
back on top.  There, done. :)

robey