dojo/uacss is leaking globals with AMD transform

[Adam Peller]

...but only after a build. It seems like whatever transform is taking place ought to insert a closure to protect the global namespace?

See console for ​http://archive.dojotoolkit.org/dojo-2010-12-16/dojotoolkit/dijit/themes/themeTester.html

[ben hockey]

(In [23376]) putting back a closure which should not have been removed in r23032. !strict refs #12098

[ben hockey]

i found a few more - will checkin after testing a build.

[ben hockey]

(In [23377]) putting back some more closures which should not have been removed in r23032. !strict fixes #12098

[Adam Peller]

Ben, these closures you added are already in functions. Shouldn't the closure protection be implicit by the AMD syntax (and provided by the transform here?) There's no reason why a var d; inside the function definition should be a global or require that additional closure.

[ben hockey]

Replying to peller:

Ben, these closures you added are already in functions. Shouldn't the closure protection be implicit by the AMD syntax (and provided by the transform here?) There's no reason why a var d; inside the function definition should be a global or require that additional closure.

i agree - we are in "limbo" until we have a fully implemented AMD loader and build system. the build transform is basically a regex to strip out the define call (and the closure created by those callback functions). so, that's the reason they still need to be in closures. there are also a number of other implications which are not ideal - #11919 covers a few more of these limitations. the current situation is definitely fragile and so i'm hoping we'll go all the way with a build tool and loader sooner rather than later.

fwiw, the changes i made just put back closures which existed before and with the current build transform, they still need to exist.

[Rawld Gill]

Replying to neonstalwart:

Replying to peller:

Ben, these closures you added are already in functions.

[snip]

i agree - we are in "limbo" until we have a fully implemented AMD loader and build system. the build transform is basically a regex to strip out the define call (and the closure created by those callback functions).

Ben--Thanks for catching/fixing this so quickly!!

I agree with your analysis. The "optimization" I made will not work after the reverse transform is applied by the current build system. fwiw...I'm nearing completion of the new build system that will resolve all these problems.