configure: Set the correct x32 machine option for yasm

[luke-jr]

No description provided.

[astiob]

Thanks! I’m afraid none of us have a x32 setup to test this (anyone correct me if I’m wrong), but if your patch does genuinely fix the build for you, I’ll accept it as is.

I have to ask though: have you tested the resulting binary to make sure it not only compiles but also actually works? I have some doubts regarding the interaction between the C pointer and integer types and the assembler code.

[luke-jr]

Best I was able to do so far was use the included test program with an image and ASS file. That did not crash, and reported something like 2 overlays merged (indicating it used assembly code, I think?). Anything else I could test with easily (ie, without needing a bunch of other codecs that don't work on x32 yet..)?

[astiob]

I’ll whip up a test ASS file and a comparison image in a minute.

[astiob]

Here, try this ASS file: https://gist.github.com/astiob/b4da312a8ec4d66d66fd
Specify 0 (or any number below 10) for the timestamp.
The output PNG should look like this: http://i.minus.com/ik71bos8rTP3.png

Wow, sorry this took so long.

[luke-jr]

Not quite. I ended up with http://imgur.com/8QRprDw

[luke-jr]

Note: I get that same result on my regular 32-bit system too...

[astiob]

Oh, I messed up the resolution. I’m fixing the gist, but it looks like it’s probably working.

[astiob]

Gist updated, new expected image: http://i.minus.com/ibeU8qGj4JtMn3.png

[astiob]

Sorry for the trouble, and thanks again! One last thing: would you mind rebasing your patch on top of our master branch? (This way we could have it marked as “merged” on GitHub without an actual merge commit, which IMO is rather pointless for small changes like this.)

[luke-jr]

Perfect match.

Rebased.

[astiob]

Thanks! Merged.