[OpenJDK 2D-Dev] RFR: 8183978 : Remove ICU layout code from OpenJDK

[Steven R. Loomis]

( So, Phil, TestLayoutVsICU is the test I was referring to the other week
which implements a Graphics2D subclass for capturing glyphs. )

TestLayoutVsICU doesn't run against ICU dynamically, within Java or
outside.   It has no dependency on 8183978 and should be left in.


What this test *does* do is to layout some text, and compare the resulting
glyphs against an XML file containing the expected glyph results.  "ICU"
here refers to a C++ test case in ICU which produced the original XML
content.  The main problems with running this test manually are, 1.) it's
font dependent— thus it is dependent on having a specific version of a font
available (hence the checksums).  If the test case were include the font,
now you have two problems, namely, legalities of redistributing the font.
  2.) there are variations with where the glyphs are located exactly, and
known differences between icu and harfbuzz as far as filler glyphs go.
(Perhaps it's a job for machine learning!)

In short, it does produce useful results, but they need human checking
right now. For example, it is useful to see especially if there is a
difference in *which* glyphs are produced.



On Fri, Oct 13, 2017 at 8:42 AM, Phil Race <philip.race at oracle.com> wrote:

> I am not sure that the test is any less valid .. it has recorded specific
> results of layout with ICU
> and runs harfbuzz comparing it to the archived results.
>
> But even as a manual test it will be a no-op without the right font
> environment.
>
> But I'll leave this up to Steven to decide as he provided us with these
> tests.
>
> One other issue with this test is that it uses JAXP classes which are
> slated to be removed ..
> Perhaps jtreg will still provide a way to access these classes on the
> classpath without
> test modifications  ???
>
> -phil.
>
>
> On 10/12/2017 02:22 PM, Sergey Bylokhov wrote:
>
>> Looks fine.
>> I guess we should update TestLayoutVsICU.java test which was added as
>> part of harfbuzz integration. I assumed that it should compares behavior
>> against icu, but it seems it is never executed.
>>
>> On 12/10/2017 10:45, Steven R. Loomis wrote:
>>
>>> approve
>>>
>>> On Wed, Oct 11, 2017 at 9:59 AM, Phil Race <philip.race at oracle.com
>>> <mailto:philip.race at oracle.com>> wrote:
>>>
>>>     Bug: https://bugs.openjdk.java.net/browse/JDK-8183978
>>>     <https://bugs.openjdk.java.net/browse/JDK-8183978>
>>>     Webrev: http://cr.openjdk.java.net/~prr/8183978/
>>>     <http://cr.openjdk.java.net/~prr/8183978/>
>>>
>>>     This fix removes the obsolete ICU opentype layout code from JDK 10.
>>>
>>>     I've built this on all platforms (including open+closed builds).
>>>
>>>     -phil.
>>>
>>>
>>>
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/2d-dev/attachments/20171013/2b2ff32a/attachment.html>