Why was a kernel-2.6.34 pushed to updates that had un-addressed bugs.

Dennis J. dennisml at conversis.de
Thu Sep 2 13:23:38 UTC 2010


On 09/02/2010 02:39 PM, drago01 wrote:
> On Thu, Sep 2, 2010 at 2:27 PM, cornel panceac<cpanceac at gmail.com>  wrote:
>>
>>
>> 2010/9/2 drago01<drago01 at gmail.com>
>>>
>>> On Thu, Sep 2, 2010 at 2:17 PM, Dennis J.<dennisml at conversis.de>  wrote:
>>>
>>>> 2. Regressions can be easier to fix because you have a "known to work"
>>>> case
>>>> you can use as a comparison. If bugs could be flagged as regression then
>>>> developers you potentially look at these first right after the
>>>> regressions
>>>> occurred and probably identify the reason for the regression right away.
>>>
>>> It isn't that easy as you make it sound (especially for the kernel).
>>> It can up to need a git bisect but that requires being able to
>>> reproduce said bug (which might require hardware that the maintainer
>>> does not have).
>>>
>>>
>> that's one of the many reasons testers' work should not just be discarded.
>
> Where did I say that?
>
>> they have a lot of hardware and a lot of time the developers can not
>> possibly have. also they are more significant as average users since they
>> are not special persons working for special companies. i assumed here that
>> the average user is important, at least as important as a(ny) company.
>
> Well yeah if a tester actually takes the time and run a bisect and
> tells the developer "this bug is caused by commit foo" it would indeed
> be very helpful.
>
> I was just replying to the statement "it is a regression and thus
> easier to fix", which isn't that simply in real world.

Just to clarify I didn't say that a regression *is* easier to fix but that 
it *can be* easier to fix due to the fact that a regression can be flagged 
as such, be pushed to the front of the queue sooner if only for a cursory 
inspection/early triaging by the developer and potentially be resolved 
because said developer still has a recent memory of the changes made.

You'll not be able to solve the problem of dealing with regressions 
completely but you can at least try to improve the process.

Regards,
   Dennis


More information about the test mailing list