RFR: 8269806: Emoji rendering on Linux [v16]

[Phil Race]

On Fri, 2 Dec 2022 19:27:30 GMT, Phil Race <prr at openjdk.org> wrote:

>> Nikita Gubarkov has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   Fix pointer to jlong conversion on x86
>
> I imported the changes and submitted to our CI test system. There seem to be a number of failures of the provided tests.
> java/awt/font/MacEmoji.java : fails on x64 and aarch64 with Error: can't find Emoji in test directory or libraries
> 
> java/awt/font/EmojiVariation.java :
> failed on linux-aarch64 with Exception: java.lang.Error: Required font not found: Noto Color Emoji
> 
> I don't think a test should fail just because there are no Emoji fonts installed.
> Lots of system configs will run into this relatively bleeding edge reqt.
> 
> and windows-64 fails this test with a REALLY long message repeating lots of times a message like
> 0023-COLOR: Expected color but rendered mono
> 
> java/awt/font/ComplexEmoji.java failed on linux-aarch64 with Exception: java.lang.Error: Required font not found: Noto Color Emoji
> 
> and failed on windows-x64 with Exception: 3: Empty image
> 
> java/awt/font/OutlineTextRendererEmoji.java: failed on aarch64 with Exception: java.lang.Error: Required font not found: Noto Color Emoji 
> 
> I haven't yet looked at how this behaves when manually testing on current platforms. These are just automated tests.

> @prrace just a few comments regarding your CI results:
> 
> 1. I don't quite understand the error message for `MacEmoji`, but note that this test was renamed to `Emoji`

Indeed the contents are a class called Emoji
I don't know why the patch still called it MacEmoji

Hmm I applied the patch skara generated and it looks like macOS's patch didn't do what it was asked

This is what was in the patch (diff) file -
diff --git a/test/jdk/java/awt/font/MacEmoji.java b/test/jdk/java/awt/font/Emoji.java
similarity index 91%
rename from test/jdk/java/awt/font/MacEmoji.java
rename to test/jdk/java/awt/font/Emoji.java
index 198db5c107fb..b44c68db7caa 100644
--- a/test/jdk/java/awt/font/MacEmoji.java
+++ b/test/jdk/java/awt/font/Emoji.java
 
> 2. All Linux fails come from absent emoji font, it would probably work with installed font

Yes, I expect that too, but the problem is we don't control all the systems used to test.

> 3. Windows renders only monochrome emoji - may it be an old Windows where there's no color emoji font? (Win8 or older?)

Not exactly its a server version of Windows 10. Today, all of the "headless" tests run on Windows Server.
Either the font isn't available for that OS, or quite possibly installing all the fonts isn't something that is done.
The test does need to be hardened against an absence of a font. Marking it headful isn't a guarantee either because they could still run on Windows Server.


> 
> So macOS seems to work and others may be just configuration problems, definitely worth trying to run manually on modern Windows & Linux.

This is largely outside of our control. They need to pass on supported configs.

> 
> And about failing from not found font: I will fix this. AFAIK currently if we gonna skip the test, we just pretend like it succeeded. There's no way to explicitly mark test as skipped, right?

Usually yes.

-------------

PR: https://git.openjdk.org/jdk/pull/4798