<Swing Dev> [9] RFR JDK-8159068:The rendering of JTable is broken

[Prasanta Sadhukhan]

On 6/24/2016 1:41 PM, Alexandr Scherbatiy wrote:
> On 6/24/2016 10:04 AM, Prasanta Sadhukhan wrote:
>> Hi Alexandr,
>>
>> Thanks for your suggestion. Have modified as per the comment to 
>> decide based on selected row. 8081491 
>> <https://bugs.openjdk.java.net/browse/JDK-8081491> testcase and 
>> SwingSet2 JTable demo works.
>> Please review modified webrev:
>> http://cr.openjdk.java.net/~psadhukhan/8159068/webrev.01/
>     - createUI() also should be called from EDT because it  uses Swing 
> components
>     - It is better to use CountDownLatch for manual tests 
> synchronization (it has special await(...) method with timeout parameter).
>
Modified testcase to use CountDownLatch
http://cr.openjdk.java.net/~psadhukhan/8159068/webrev.02/

Regards
Prasanta
>   Thanks,
>   Alexandr.
>>
>> Regards
>> Prasanta
>> On 6/23/2016 12:41 PM, Alexandr Scherbatiy wrote:
>>> On 6/23/2016 9:53 AM, Prasanta Sadhukhan wrote:
>>>>
>>>>
>>>> On 6/23/2016 12:10 PM, Alexandr Scherbatiy wrote:
>>>>> On 6/21/2016 2:57 PM, Prasanta Sadhukhan wrote:
>>>>>>
>>>>>> On 6/21/2016 5:16 PM, Alexandr Scherbatiy wrote:
>>>>>>> On 6/21/2016 1:58 PM, Prasanta Sadhukhan wrote:
>>>>>>>>
>>>>>>>> On 6/21/2016 4:14 PM, Alexandr Scherbatiy wrote:
>>>>>>>>> On 6/20/2016 8:10 AM, prasanta sadhukhan wrote:
>>>>>>>>>>
>>>>>>>>>> Gentle reminder for review!!
>>>>>>>>>>
>>>>>>>>>> Regards
>>>>>>>>>> Prasanta
>>>>>>>>>> On 6/13/2016 4:31 PM, prasanta sadhukhan wrote:
>>>>>>>>>>> On 6/13/2016 12:51 PM, prasanta sadhukhan wrote:
>>>>>>>>>>>> Hi All,
>>>>>>>>>>>>
>>>>>>>>>>>> Please review a fix for jdk9 where it was seen that if we 
>>>>>>>>>>>> try to select some rows in a JTable, the text painted in 
>>>>>>>>>>>> the rows goes missing.
>>>>>>>>>>>>
>>>>>>>>>>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8159068
>>>>>>>>>>>>
>>>>>>>>>>>> webrev: 
>>>>>>>>>>>> http://cr.openjdk.java.net/~psadhukhan/8159068/webrev.00/
>>>>>>>>>>>>
>>>>>>>>>>>> The issue was rMax value was decremented wrongly so when 
>>>>>>>>>>>> paintCells() is called with wrong rMax, some rows were not 
>>>>>>>>>>>> printed correctly.
>>>>>>>>>>>>
>>>>>>>>>>>> Fix is to make sure rmax is decremented properly, only when 
>>>>>>>>>>>> we are trying to print whole visible portion of JTable and 
>>>>>>>>>>>> NOT when some rows are being painted.
>>>>>>>>>
>>>>>>>>>   Could you give two samples how this algorithm work. One 
>>>>>>>>> sample where a whole visible portion of a JTable and another 
>>>>>>>>> where some rows are being printed. What are rMax and  rMin 
>>>>>>>>> values in both cases and how are they calculated?
>>>>>>>>>
>>>>>>>> If a JTable is of 50 rows and only 35 are being visible in page 
>>>>>>>> 1, then
>>>>>>>> if whole visible portion of JTable is printed, rMin will be 0 
>>>>>>>> and rMax was 35
>>>>>>>> so 36 rows were getting printed so I decrement rMax by 1 to 34 
>>>>>>>> so only 35 will be printed (same as shown on console).
>>>>>>>> When we select some row of JTable as in the case of LostText 
>>>>>>>> testcase, rMin will be say 6 and rMax will be 9 in which case 
>>>>>>>> also, I was decrementing rMax so rMin=6, rMax=8 so next row was 
>>>>>>>> not getting painted.
>>>>>>>    And what are indices of the selected rows?
>>>>>>>
>>>>>> It will depend on the last selection. At start, rMin = 0 , rMax = 
>>>>>> last indice, say 10 for a JTable of 10 rows
>>>>>> Now, if we select row 5, rMin and rMax both becomes 5 and we 
>>>>>> decrement rMax so rMax becomes less than rMin and paintCell() due 
>>>>>> to this check
>>>>>> (int row = rMin; row <= rMax; row++) it does not do
>>>>>> paintCell(g, cellRect, row, column)
>>>>>> and nothing gets painted.
>>>>>    It is not clear how the selection interval [0..N] is 
>>>>> distinguishable from the case where is no selection because rMin 
>>>>> should be 0 in both case. May be it is better to decrement the 
>>>>> rMax depending on are there selected rows or not.
>>>> Can you suggest if there any way we can find out if there are any 
>>>> selected rows or not?
>>>    There is the method table.getSelectedRowCount(). May be for 
>>> performance reason it is better to check table.getSelectedRow().
>>>
>>>   Thanks,
>>>   Alexandr.
>>>
>>>>
>>>> Regards
>>>> Prasanta
>>>>>
>>>>>   Thanks,
>>>>>   Alexandr.
>>>>>>
>>>>>> Regards
>>>>>> Prasanta
>>>>>>>   Thanks,
>>>>>>>   Alexandr.
>>>>>>>
>>>>>>>>
>>>>>>>> Regards
>>>>>>>> Prasanta
>>>>>>>>>   Thanks,
>>>>>>>>>   Alexandr.
>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> Regarding the regression testcase, I could not make it 
>>>>>>>>>>>> automated as the failure happens on random iteration.
>>>>>>>>>>>> and also, getting selection background/foreground was 
>>>>>>>>>>>> giving same values with and without the missing text.
>>>>>>>>>>>>
>>>>>>>>>>> Also, since it is a regression of 8081491 
>>>>>>>>>>> <https://bugs.openjdk.java.net/browse/JDK-8081491>, it's 
>>>>>>>>>>> testcase are working fine with this fix and so did SwingSet2 
>>>>>>>>>>> JTable demo.
>>>>>>>>>>>> Regards
>>>>>>>>>>>> Prasanat
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/swing-dev/attachments/20160624/d27a4d78/attachment.html>