[OpenJDK 2D-Dev] RFR: 8132850: java.lang.ArrayIndexOutOfBoundsException during text rendering with many fonts installed

[Phil Race]

Hi all,

Still looking for any reviewers for this 7 day old review ...

-phil.

On 8/5/15 12:15 PM, Phil Race wrote:
> https://bugs.openjdk.java.net/browse/JDK-8132850
> http://cr.openjdk.java.net/~prr/8132850/
>
> The bug is an ArrayIndexOutofBoundsException when a very large set of 
> fonts are installed on Linux.
> It occurred rendering from a compositefont :
>
> Exception in thread "AWT-EventQueue-0" 
> java.lang.ArrayIndexOutOfBoundsException: -35127285
>     at sun.font.FileFontStrike.setCachedGlyphPtr(FileFontStrike.java:520)
>     at sun.font.FileFontStrike.setCachedGlyphPtr(FileFontStrike.java:537)
>     at 
> sun.font.FileFontStrike.getSlot0GlyphImagePtrs(FileFontStrike.java:449)
>     at 
> sun.font.CompositeStrike.getGlyphImagePtrs(CompositeStrike.java:115)
>     at sun.font.GlyphList.mapChars(GlyphList.java:272)
>
> Note that you do not see the exception on JDK 9 because it is caught 
> as a  result of
> some other work, but it is seen on 8, and the underlying problems 
> affect all releases.
>
> There are several contributing problems :
>
> - the native fontconfig code to accumulate the fonts was resetting the 
> set of glyphs for each time through the loop.
>  thus its heuristic to limit the number of slots was never met and so 
> it would add more slots than it should.
>  In addition it should limit to 256 (or fewer) anyway.
>
> - a special case in FileFontStrike accepted a composite glyph id. 
> These use the MS byte of the int to store
> a slot. Once it was > 127 then the test to see if the glyph was a slot 
> zero glyph failed. Hence the AIIOB.
>
> - finally for extra safety the CompositeFont constructor should ignore 
> trailing slots that will not fit.
>
> I could not devise a regression test since its requires a custom 
> system config
> and in any case the exception is not seen on 9 as noted above.
>
> -phil.