Memory leak on constructor error

[elihart]

When I try to create a new GifDrawable(byte[]) and it fails because the file is not a gif, I am left with a dangling byte[] of the size of the image that can never be freed. Over time this leads to large memory problems.

The problem seems to be somewhere in openByteArray(), but I haven't been able to track down the exact problem in the C code.

I load lots of images that may or may not be gifs, and I was using this as an easy test to check if they are gifs. Unfortunately the memory leak is prohibitive. I'd love some input on where to look to fix this.

[koral--]

Thanks, you are right, some memory cleanings are missing in case when opening fails. I will fix it ASAP. If you want to it yourself look at Java_pl_droidsonroids_gif_GifDrawable_openByteArray in jni/gif.c.
Before returning NULL we need to DeleteGlobalRef container->buffer and free the container like in Java_pl_droidsonroids_gif_GifDrawable_free (which is called normally from Java on cleaning up normally opened GIFs). Other functions like Java_pl_droidsonroids_gif_GifDrawable_open<type> need to be checked if there are no similar bugs inside.
Pull request will be appreciated.

[elihart]

Ok, good to know I'm not going crazy. Tracking down that leak was tricky.

I was already looking in Java_pl_droidsonroids_gif_GifDrawable_openByteArray but I haven't worked with C in a while and I wasn't able to find the right pointers to free.

Based on your advice I tried replacing

return open(GifFileIn, Error, container->pos, byteArrayRewindFun,
            env, metaData);

with

jint openResult = open(GifFileIn, Error, container->pos, byteArrayRewindFun,
            env, metaData);

    if (openResult == NULL) {
        (*env)->DeleteGlobalRef(env, container->buffer);
        free(container);
    }

    return openResult;

but that didn't work. I'd love to help out and submit a pull request if I can get it fixed, but unfortunately I'm struggling with C at the moment.

[koral--]

Did you recompile native sources properly?

[elihart]

No, I didn't! I haven't worked with jni before and wasn't aware I had to recompile it manually.

Anyway, I recompiled and now it seems to be working, thanks!

[koral--]

Fixed for all affected sources.