[4.0-pre] parsing invalid input

[alexlamsl]

With the latest version, <![CDATA[p.b{background:red}]]> get parsed and gives <![CDATA[p.b{background:red} whereas 3.x will leave it as is.

Ideally it should throw an error, but in any case html-minifier has already dropped support for CDATA, so this isn't the end of the world but I thought I'll let you know of the behavioural difference.

[jakubpawlowicz]

Thanks @alexlamsl - it should definitely throw an error about invalid input.

[alexlamsl]

There is also *{font: "monospace"} which transforms into *{font:&quot monospace&#34} without any errors.

(The HTML entity characters are intentionally left not decoded.)

[alexlamsl]

As of a239f68, *{font: "monospace"} now works (fails) the same way as 3.4.x

One down, one to go 👍

[jakubpawlowicz]

Thanks @alexlamsl for keeping a close eye on this!

[jakubpawlowicz]

I've added more rules to ignoring invalid selectors but that list may not be exhaustive.

[alexlamsl]

If I got this right, the current master <![CDATA[\np.a{background:red}\n]]> minifies to an empty string. In previous versions it would only replace those \n to whitespaces.

Also, this test now fails:

p.h{background:red}<!--
p.i{background:red}
-->p.j{background:red}

Expects

p.h{background:red}<!-- p.i{background:red}-->p.j{background:red}

Actual

p.h{background:red}-->p.j{background:red}

[alexlamsl]

There isn't anything in the errors array either, so I can't really workaround these cases easily. 😕

[alexlamsl]

Ah it's being reported as a warning instead. Skipping minification on sighting on any warnings seem a little bit of an overkill though?

[alexlamsl]

I've put up a branch for illustrating the failures: https://github.com/kangax/html-minifier/tree/clean-css-4

After checking out and switching to the branch, just npm install then npm test to see the failures.

[jakubpawlowicz]

@alexlamsl yup, these are warnings for the reason you mentioned. If browsers safely ignore such rules so should we.

Thanks for the clean-css-4 branch. Will check out those failures.

[alexlamsl]

@jakubpawlowicz so if I understand you correctly, with those chunks chopped off there won't be any loss of (CSS) functionality (read: broken the same way) in modern web browsers?

If so, I guess html-minifier can just work with that - after all, appearing the same way after minification in web browsers is the end goal. 😎