[11u] java/awt/FontMetrics/MaxAdvanceIsMax.java test failure (was: [11u] RFR 8210782: Upgrade HarfBuzz to the latest 2.3.1)

[Lindenmaier, Goetz]

Hi Martin,

Yesterday I filed 
8223869: Problem list java/awt/FontMetrics/MaxAdvanceIsMax.java on more platforms
as I found it is excluded on solaris and mac already. 

Thanks for the explanation.  
I will look at what freetype libs are used on the systems where this is failing.
Maybe we should mark the test with @key intermittent if it depends
on the system setup that much?

Best regards,
  Goetz.


> -----Original Message-----
> From: Martin Balao <mbalao at redhat.com>
> Sent: Tuesday, May 14, 2019 6:41 PM
> To: Lindenmaier, Goetz <goetz.lindenmaier at sap.com>; 'Severin Gehwolf'
> <sgehwolf at redhat.com>; Langer, Christoph <christoph.langer at sap.com>;
> jdk-updates-dev at openjdk.java.net
> Cc: 2d-dev <2d-dev at openjdk.java.net>; build-dev at openjdk.java.net;
> Martin Balao Alonso <mbalaoal at redhat.com>
> Subject: Re: [11u] java/awt/FontMetrics/MaxAdvanceIsMax.java test failure
> (was: [11u] RFR 8210782: Upgrade HarfBuzz to the latest 2.3.1)
> 
> Hi Goetz,
> 
> On 5/13/19 1:38 PM, Lindenmaier, Goetz wrote:
> >
> > Can I somehow verify that it's the font that has the problem?
> > Can I fix the font so that the test passes?
> >
> 
> I cannot say whether or not the static max advance value in each font is
> right or not, but let's assume it is. The underlying problem here is
> that OpenJDK uses a couple of internal FreeType library values to
> calculate the effects of algorithmic bold and italic in the max advance
> value -"algorithmic" means that the font is not italic or bold and is up
> to the rendering engine to generate the desired effect-. These values
> have changed over the years. What we did in 8218854 [1] was updating the
> italic value to the latest version and supporting bold in the
> calculation. Ideally, OpenJDK should not be tight to these values.
> However, that's not easy to get rid of unless we change the API or the
> API semantics. All this means that if you use OpenJDK 11 with an old
> Free Type library, you may have different values and the test may fail.
> 
> Note: this is just an hypothesis, I couldn't reproduce on my own.
> 
> I believe the test assertion is right if we consider API semantics only
> but we can put some constraints given reality.
> 
> Kind regards,
> Martin.-
> 
> --
> [1] - http://hg.openjdk.java.net/jdk/jdk/rev/0804f29e8be7
>