Non-deferred scripts may block with the HTML parser, we should defer them where possible.

Comment on attachment 808645 [details] Github pull request Hey - could one of you guys review this quick patch? Thanks!

Hi Kevin, this is a bit weird for me, the scripts already include the |defer| attribute, is there any bug that prevent the script to be deferred for having |type| attribute? Maybe I'm missing something :S

(In reply to Francisco Jordano [:arcturus] from comment #3) > Hi Kevin, > > this is a bit weird for me, the scripts already include the |defer| > attribute, is there any bug that prevent the script to be deferred for > having |type| attribute? > > Maybe I'm missing something :S Hi Francisco, Sorry or the confusing pull request! The defer attribute was added to l10n and l10n_date libraries. The type attribute was removed becuase it is not needed. These are two conventions that we use in other Gaia apps. This should result in a tiny perf boost. Let me know of you have any other questions, thanks!

Comment on attachment 808645 [details] Github pull request Thanks Kevin!

Thanks for the quick reviews! Landed in master: https://github.com/mozilla-b2g/gaia/commit/30115b14062c39443528ab8c9a79e7f424178e36

For what its worth, I was concerned that this might have caused an extra reflow, so I took a couple profiles: Before: http://people.mozilla.org/~bgirard/cleopatra/#report=0bfe988bac496ba7a7e09f8cc2e3623099ca0afe After: http://people.mozilla.org/~bgirard/cleopatra/#report=cc2f0dacb5a5f188c21ea1b2d08631dfd98190e2 I don't see any reflow regression. As expected the static loading in html5::RunFlushLoop() improved from 400+ms to ~350ms. Just including the profiles here in case anyone wants to look at them.

Thanks for the profiles Ben!

Awesome :) Thanks for the info Ben

Uplifted 30115b14062c39443528ab8c9a79e7f424178e36 to: v1.2: 96f48b5e0acc8bee3303f0569b699494b06586ab