<Swing Dev> [10] RFR JDK-8191428: Regression: Swing button label wrapping with hidpi

Prasanta Sadhukhan prasanta.sadhukhan at oracle.com
Tue Nov 21 09:51:28 UTC 2017

Hi Sergey,

I created the automated test testing the fact that if text is broken 
into multiple lines, then button rendering the text will have its height 
increased also.

On 11/21/2017 6:49 AM, Sergey Bylokhov wrote:
> Hi, Prasanta.
> Is it possible to create an automated test for this issue?
> (I think that we were just lucky that Phil found this problem by the 
> demo)
> On 20/11/2017 03:26, Prasanta Sadhukhan wrote:
>> Hi All,
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8191428
>> It is seen that with
>> 8172025: HiDPI with non-integer scale factor - SPANs in HTML are 
>> rendered overlapping each other
>> fix, the html text in button is rendered in different lines even 
>> though it is specified as a one word.
>> This is because with getSpan() returning float value (say 29.9996)
>> FlowView.layoutRow() calculates "spanLeft" using getFlowSpan() method 
>> which uses "int" calculation (so get 29]
>> whereas "chunkSpan" got from GlyphView#getTabbedPane (using 
>> GlyphPainter1.getSpan) gets "float" value [ie 29.9996]
>> so this calculation (chunkSpan > spanLeft) is satisfied  [ex, 29.996 
>> > 29] and so it calls breakView() to break the text.
>> In the proposed fix, I have reverted the getSpan() value to return 
>> "int" so as to prevent this regression.
>> This fix also does not cause any problem with 8172025 issue which 
>> renders fine.
>> webrev: 
>> http://cr.openjdk.java.net/~psadhukhan/8191428/webrev-1stapproach/webrev.00/ 
>> Another approach is to use "float" calculation for all "spans" 
>> calculation but that will cause many changes, including changing 
>> public/protected method signature
>> which I believe is risky at this point of time for jdk10, as it might 
>> potentially cause more regressions.
>> Sneak peek of 2nd approach changes are:  [work in progress]
>> http://cr.openjdk.java.net/~psadhukhan/8191428/webrev-2ndapproach/
>> Regards
>> Prasanta

More information about the swing-dev mailing list