On 2009/08/09 19:30:28, jat wrote:
> I haven't taken a look at this closely (I had read Aaron Boodman's
blog about
> the approach at

http://www.aaronboodman.com/2009/07/bundling-multiple-versions-of-binary.html
> previously), but I am not sure this can detect all the cases we need
to
> differentiate.

Thanks for the link. I hadn't read the blog post.

> For example, on Fedora Core 10 FF3.0.11, the plugin needs to be linked
against
> different libraries (the ff3+ entry there now), and it isn't clear how
we can
> differentiate that in the JS stub from other FF3.0.11 builds that
require the
> ff3 libraries.

We can still build a separate XPI for 64-bit FC 10 by invoking Make with
the
appropriate arguments. We can also include the libraries for all
versions of
Firefox for that platform in that XPI.

> Also, given the other things that we need to get done for GWT 2.0 MS1,
I am
> leery of making such a major change that might distract from getting
those
> things done.

Okay, but the changes are not as big as they might seem from the CL.
Mostly
it's just deleting unused files.  The modifications and additions are
pretty
much:

components/stub.js: the loader from Gears

Makefile: outputs libraries in lib/<platform>/<ff-version> instead
   platform/<platform>.

config.mk: one-line change to always build universal binaries on the mac


> So, my inclination would be to wait on this until after all the other
> OOPHM-related pieces are done for MS1 and if there is time left then
we can
> pursue it further.

What are the OOPHM-related goals for MS1?

I've found that having a single XPI made making changes to OOPHM much
easier,
since you can install a single extension and test on all versions of
Firefox.

That said, if you want to hold off on this, I can understand.

http://gwt-code-reviews.appspot.com/56808

--~--~---------~--~----~------------~-------~--~----~
http://groups.google.com/group/Google-Web-Toolkit-Contributors
-~----------~----~----~----~------~----~------~--~---

Reply via email to