[IMAGING-107] sanselan throws a java.lang.NegativeArraySizeException on at least this one .jpg file

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 0.97
• Fix Version/s: 1.0-alpha1
• Component/s: Format: JPEG

Environment

Linux, Java(TM) SE Runtime Environment (build 1.7.0_21-b11)
(though I doubt that matters.)

• Assignee: Unassigned
• Reporter: Doug McLaren
• Created: 03/Jun/13 20:58
• Updated: 07/Aug/20 04:32
• Resolved: 23/Jun/13 21:43

Description

Saneslan seems unable to run getImageInfo() on the attached file, gives this error –

java.lang.NegativeArraySizeException
at org.apache.sanselan.common.BinaryFileFunctions.readByteArray(BinaryFileFunctions.java:496)
at org.apache.sanselan.formats.jpeg.JpegUtils.traverseJFIF(JpegUtils.java:94)
at org.apache.sanselan.formats.jpeg.JpegImageParser.readSegments(JpegImageParser.java:170)
at org.apache.sanselan.formats.jpeg.JpegImageParser.readSegments(JpegImageParser.java:267)
at org.apache.sanselan.formats.jpeg.JpegImageParser.getImageInfo(JpegImageParser.java:659)
at org.apache.sanselan.Sanselan.getImageInfo(Sanselan.java:589)
at org.apache.sanselan.Sanselan.getImageInfo(Sanselan.java:563)
at org.apache.sanselan.Sanselan.getImageInfo(Sanselan.java:582)
at SanselanTest.main(SanselanTest.java:68)

Now, this error seems very similar to IMAGING-21, but it's not quite the same for two reasons –

1) the stack is somewhat different, and
2) the fix provided for that bug does not fix this issue.

Now, looking at the file that triggers this issue, other tools throw a warning for it –

gimp says: Corrupt JPEG data: 2715 extraneous bytes before marker 0xe2
xv says the same thing.
convert from ImageMagick converts it with no error or warning.
jhead says: Nonfatal Error : 'pod_tile_foundation.jpg' Extraneous 2715 padding bytes before section E2
jpegtran says: Corrupt JPEG data: 2715 extraneous bytes before marker 0xe2

... so it seems there's something actually wrong with this file, but even so, the image is there, and other tools can determine the dimensions of it, so in theory sanselan ought to be able to as well.

I'll include the somewhat corrupt file that is triggering this exception, as well as the test case program (which came straight from IMAGING-21.)

The test case also checks the two files that were included in IMAGING-21 just to show that that problem was fixed (at least once I applied the fix included there.)