RFS: 0ad

[Philip Taylor]

On Sun, Apr 10, 2011 at 11:04 AM, Paul Wise <pabs at debian.org> wrote:
> On Sun, Apr 10, 2011 at 5:49 PM, Vincent Cheng <vincentc1208 at gmail.com> wrote:
>
>> Oops, don't know why I broke off mid-sentence like that. My point was that
>> porting 0 A.D. to work with newer versions of Spidermonkey seems to be a lot
>> of work for very little gain and lots of opportunities for potential
>> breakage. Philip addressed this already in an earlier message [1].

To clarify slightly: I'm happy to port to SpiderMonkey 1.8.5, which is
the newest release (two weeks old). (The version numbers in e.g.
'libmozjs-dev 1.9.1.18' are Gecko version numbers, not SpiderMonkey
version numbers, and don't correspond to any official standalone
SpiderMonkey release, so they're older than 1.8.5). If there are
occasional further standalone releases of SpiderMonkey in the future,
I expect I'll be happy to port to them too.

The main constraint is that any given version of the game will only
support a single SpiderMonkey release (and will use that same release
across all Linux distros and all other platforms). If we use something
like libmozjs-dev 1.9.1.18, there's no guarantee that libmozjs-dev
1.9.1.19 won't break API compatibility or cause multiplayer sync
errors, so I don't want to use a version like that. But if we support
the js185-1.0.0.tar.gz release then a hypothetical bugfixed
js185-1.0.1.tar.gz should work too (that stability being the purpose
of the standalone releases), and we can stick with that until there's
a new standalone release series that we can port to (or until it
becomes apparent there won't be any more standalone releases and then
we can revisit the problem), so it should be less of a maintenance
burden than the alternatives.

> Maybe porting to another JavaScript engine (like Google V8) is the
> best long-term solution?

I believe that will never happen - it'd be a huge amount of work for
very little benefit (and that's assuming V8 does provide a stable API
that they will never break in order to make Chrome 5% faster - I don't
know if that's the case).

-- 
Philip Taylor
excors at gmail.com