+1 with John.

Renaming autotest_lib to autotest will have a hugely impact to ChromeOS test
automation project, since we need to go over through all test cases to
change the import statement...

Eric

On Tue, Jan 4, 2011 at 9:21 AM, John Admanski <[email protected]> wrote:

>
>
> On Tue, Jan 4, 2011 at 9:08 AM, Lucas Meneghel Rodrigues 
> <[email protected]>wrote:
>
>> On Tue, 2011-01-04 at 14:31 -0200, Eduardo Habkost wrote:
>> > On Tue, Jan 04, 2011 at 02:15:26PM -0200, Lucas Meneghel Rodrigues
>> wrote:
>> > > I see an opportunity for us to do some cleanup on that and eliminate
>> > > bin. Now, we need a plan of migration, a carefully made patchset, and
>> > > tons of testing. Perhaps the entire contents of bin can be dropped
>> into
>> > > client altogether, but that is not too symmetrical with common_lib.
>> >
>> > IMO there's no need to hurry. If we think it is worth it, we can slowly
>> > move libraries out of bin and keep only executable scripts, but no need
>> > to do it in a large step.
>> >
>> > About symmetry with common_lib: I think common_lib is the aberration (it
>> > could live outside 'client', but we have the issue of sending only the
>> > 'client' directory to test hosts), so I wouldn't worry about making the
>> > client-only modules layout match the common_lib layout.
>> >
>>
>> It seems to me that we can make things much better by doing some simple,
>> mechanical steps:
>>
>> 1) Change autotest_lib to autotest
>>
>
> I think that from an aesthetic point of view autotest is a nicer name than
> autotest_lib, but it does clash with server/autotest.py so if you want to
> rename the top-level package you'll need to rename that too, which is more
> painful than it sounds because server/autotest.py is a very prominent part
> of the server-side test API. It also amplifies the pain of step 4 a lot
> since it means updating just about every internal server-side control file,
> anywhere.
>
>
>> 2) Move common_lib to common
>> 3) Move stuff from client/bin to client altogether
>> 4) Telling everybody that has internal tests/site extensions to adapt
>> their tests to the changed API.
>>
>> The 3 first steps will generate *freaking huge* patchsets, but it's just
>> mechanical work. As long as it's well tested and audited (pylint comes
>> to the rescue here), it's not that dangerous.
>>
>> Now, 4) is probably the hardest part, as we have parties with large site
>> repositories, google being the biggest. So for me, is all a matter for
>> the google folks (Chrome OS and other internal users) to pronounce about
>> it. I won't even try to make a change if those guys do not feel OK about
>> it.
>>
>> Any comments guys?
>>
>> Lucas
>>
>>
>


-- 
Eric Li
李咏竹
Google Kirkland
_______________________________________________
Autotest mailing list
[email protected]
http://test.kernel.org/cgi-bin/mailman/listinfo/autotest

Reply via email to