HTMLImageElement::matches_environment doesn't look quite right.

[emilio]

It's passing ParsingMode::all(), which doesn't really seem like it wants to pass, it should probably pass DEFAULT instead.

This should be testable, I think in Servo the sizes attribute will incorrectly work with min-width: <integer> or such.

[emilio]

I added a couple FIXMEs for this in #21588

[emilio]

cc @nupurbaghel

[nupurbaghel]

So finally it has been solved then ?

[jdm]

It has not been solved.

[nupurbaghel]

I tried using min-width, max-width queries as well as different combinations of width of images in srcset and according to my understanding, what is happening for now is-

• any following src is ignored (as mentioned in algo step 4.3)
• in case any query is not matched (which will be true if we are using min-width: <integer> value eliminating the units), then the max-width image is selected from the srcset.
  The behavior is the same for both parse mode : all() & DEFAULT

As I currently do not realize well how do the 2 parse modes actually differ, any ideas on how to write a test which will give differing tests on both of these will be really beneficial.

Just for reference, the iframe below with the mentioned img tag displays a red image -

<iframe id="frame1" style="width:150px; height:150px" src="......">
</iframe>
<img srcset="/images/red-16x16.png 96w,
            /images/blue96x96.png 16w"
     sizes="(min-width: 200) 96px,
            (min-width: 100) 16px"
     src="/images/green-256x256.png">

[emilio]

So, ParsingMode::all() is effectively ALLOW_UNITLESS_LENGTH | ALLOW_ALL_NUMERIC_VALUES.

I think we're very lax with the parsing of sizes, but (min-width: 1) will produce an always-matching media query now, and should produce a never-matching media query instead.