Re: Late binding of integer constants?

Mark Raynsford <[email protected]> Thu, 16 Nov 2017 13:50:31 +0000

Newsgroups	gmane.comp.java.objectweb.asm
Organization	io7m.com
Message-ID	<[email protected]>

On 2017-11-16T14:34:53 +0100
<[email protected]> wrote:
>
> In other words, the SpHttpStatus.SC_OK constant has been inlined and replaced
> with its value. However, that constant is defined in a different module than
> the calling code, meaning that if at runtime, a version with a different value
> is used, things will break. I didn't expect that (I was expecting late
> binding) - does anyone on this list know something about the rationale for
> this?

Unless I'm mistaken, this is ancient javac behaviour. The fact that
the inlining crosses module boundaries is just coincidental, as there
was obviously no such thing as a module boundary when the behaviour
was first specified. I wrote a blog post about it a while back:

  http://blog.io7m.com/2017/06/30/mind-your-constants.xhtml

It's documented in §13.4.9 of the JLS:

  https://docs.oracle.com/javase/specs/jls/se9/html/jls-13.html#jls-13.4.9

-- 
Mark Raynsford | http://www.io7m.com

signature.asc (application/pgp-signature, 833 B)