[vtk-developers] Dashboard test failure question

[Bill Lorensen]

Operator error... Should be won't fix.


On Fri, Jan 17, 2014 at 2:24 PM, David Gobbi <david.gobbi at gmail.com> wrote:
> I understand.  However, if the bug isn't fixed and there is a very low
> probability that it every will be fixed, it should be marked as "won't fix"
> instead of being marked as "fixed".
>
> So I'm just wondering how it came to pass that the bug is marked as
> fixed, when it isn't.
>
>   David
>
>
> On Fri, Jan 17, 2014 at 12:11 PM, Bill Lorensen <bill.lorensen at gmail.com> wrote:
>> I'm pretty sure that did not fix the bug.
>>
>> We spent a lot of time on this one. The volume rendering code is old
>> and tough to follow.
>>
>>
>>
>> On Fri, Jan 17, 2014 at 1:38 PM, David Gobbi <david.gobbi at gmail.com> wrote:
>>> I went back to the bug report on this multi-threading issue, and oddly
>>> enough the bug is marked as "fixed":
>>> http://www.vtk.org/Bug/view.php?id=13420
>>>
>>> A note in the bug report says "Was fixed in commit: e4a793c47cc3"
>>> But I looked at the commit, and it seems to be unrelated.  Why was
>>> the bug closed, when the problem still exists?  Was it closed
>>> accidentally?
>>>
>>>   David
>>>
>>>
>>> On Fri, Jan 17, 2014 at 10:40 AM, David Gobbi <david.gobbi at gmail.com> wrote:
>>>> On Fri, Jan 17, 2014 at 10:07 AM, David Cole <dlrdave at aol.com> wrote:
>>>>>
>>>>> If that warning is true, then shouldn't any test that uses that class set
>>>>> the number of threads to 1 instead of trying to use all the cores available
>>>>> on a machine......?
>>>>
>>>> Bill already fixed a bunch of tests to set NumberOfThreads=1 for this
>>>> mapper.  TestSmartVolumeMapperWindowLevel must have slipped through
>>>> the cracks.
>>>>
>>>>> Does anybody really want non-repeatable results?
>>>>
>>>> Fixing the bug in the mapper would be the correct way of ensuring
>>>> repeatable results: it's a multi-threaded mapper, but running it
>>>> with too many threads causes artifacts.  In my view, the
>>>> "SetNumberOfThreads(1)" trick is a way of sweeping the problem
>>>> under the rug, out of sight and out of mind.
>>>>
>>>>   David



-- 
Unpaid intern in BillsBasement at noware dot com