Possible binary incompatibility between 0.5 and 0.7 about binary

[kolmodin]

The format is not supposed to change. I can't see any changes at all in that code for at least the last 5 years.
You're absolutely supposed to be able to encode something with 0.5 and decode with 0.7, and the other way around. Anything else is a bug.

There was a blog post discussing the new addition to the API - but no format changes.
http://lennartkolmodin.blogspot.ru/2013/03/binary-07.html

Can you narrow it down a little bit?

[nomeata]

Can you narrow it down a little bit?

Ok, I’ll do that.

[kolmodin]

Thanks.
The internals changed a lot between 0.5 and 0.7. Binary used to be a state monad and is not a CPS monad.

I think it's the getLazyByteString that has become more picky. It seems like it previously didn't fail if there wasn't enough data available, and now it does.
https://github.com/kolmodin/binary/blob/0.5.1.1/src/Data/Binary/Get.hs
https://github.com/kolmodin/binary/blob/master/src/Data/Binary/Get.hs

It's used in src/Data/MyText.hs:34 in arbtt.

[nomeata]

Oh, well spotted. That must be it – hence the comment “safe approximation”.

Given that the docs of getLazyByteString in binary-0.5.1.1 explicitly state that it does not fail if there are fewer bytes: Was the change in 0.7 deliberate?

I’ll see if replacing the code by code that takes remaining into account solves the problem.

[nomeata]

Yes, that was it; now the test suite also works with binary-0.7. Thanks a lot for the help!

I’m closing this (but I opened #45 :-))

[kolmodin]

The change was not deliberate, I did it by mistake. I like the new API better though.
This is the first time the issue has been brought up in 18 months, so I hope the impact of this "silent" change is not significant.

Sorry for the trouble and thanks for the report!