Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Indefinite loop while reading corrupt JPEG

See original GitHub issue

Hi Harald.

I know reading corrupt files and solving related problems is boring and not very interesting, but since this one hangs the thread it might be worth looking into. Tests have been performed on my code using some Google test suite, and it ends up hanging the application because the thread doing the reading goes into an indefinite loop while holding locks ending up blocking almost everything.

I’ve been able to track it down to an endless loop in com.sun.imageio.plugins.jpeg.JPEGImageReader.hasNextImage(). The loop is this:

        boolean foundFF = false;
        for (int byteval = iis.read();
             byteval != -1;
             byteval = iis.read()) {

            if (foundFF == true) {
                if (byteval == JPEG.SOI) {
                    iis.reset();
                    if (debug) {
                        System.out.println("true");
                    }
                    return true;
                }
            }
            foundFF = (byteval == 0xff) ? true : false;
        }

From what I can tell, the only ways to break the loop is either by an Exception being thrown, -1 being returned from the ImageInputStream or finding the SOE. I know that this code isn’t yours, but it seems that the reason for the endless loop is found in com.twelvemonkeys.imageio.plugins.jpeg.JPEGSegmentImageInputStream.repositionAsNecessary() which swallows the EOFException. The code comments says that an EOFException will be thrown or -1 will be returned by the subsequent read, but this clearly isn’t the case.

I’m not familiar enough with this code to understand why the EOFException is ignored in the first place, and thus not able to produce a fix.

The test suite has thousands of images, and only 8 JPEGs end up in a hung thread, so it’s not a very frequent problem. Still it would be nice to avoid it.

The corrupt/invalid JPEGs are very small (as they have very little actual content), so I’m attaching them all:

id_005573 sync_jpeg_turbo src_004235 id_000448 src_000000 op_havoc rep_32 cov id_001917 src_001912 op_havoc rep_8 id_001918 src_001917 op_havoc rep_8 id_003682 sync_jpeg_turbo src_001907 id_003723 sync_jpeg_turbo src_001917 id_003809 sync_jpeg_turbo src_001918 id_004235 src_003897 op_havoc rep_16