You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Jetpack has a new lazy loading images feature that serves tiny 1x1 placeholders for all images, and then swaps in (and loads) the real src URLs when they're scrolled into view.
fortunately it also serves duplicate <img>s with the real URLs inside <noscript>s, which means mf2 consumers like bridgy still work...except every u-photo gets duplicated, one with the real image and one with the jetpack lazy placeholder. eg HTML like this from http://paulatlan.net/things-that-surprise-me/petits-bonheurs/ :
discovered an interesting meaningful difference in mf2py's output based on the underlying html parser: html5lib omits the contents of <noscript> tags; lxml includes them.
we use lxml in app engine prod, but only by default. we were actually using html5lib locally. so when i wrote the test for this fix, it initially failed because it didn't include the full <img> from inside the <noscript>. i'm switching it to use lxml explicitly, and pin the same version as app engine, to prevent that skew.
Jetpack has a new lazy loading images feature that serves tiny 1x1 placeholders for all images, and then swaps in (and loads) the real
src
URLs when they're scrolled into view.fortunately it also serves duplicate
<img>
s with the real URLs inside<noscript>
s, which means mf2 consumers like bridgy still work...except everyu-photo
gets duplicated, one with the real image and one with the jetpack lazy placeholder. eg HTML like this from http://paulatlan.net/things-that-surprise-me/petits-bonheurs/ :results in this mf2:
we should probably just special case and ignore all
*/lazy-images/images/1x1.trans.gif
images.The text was updated successfully, but these errors were encountered: