Search code examples
cssgoogle-chromeflexboxcss-specificity

Did Chrome 40 break justify-content CSS overriding?

I noticed with today's Chrome 40 update that justify-content does not seem to get properly overriden by subsequent style declarations.

See this fiddle for an example:

  <div class="flex-parent">
    <div class="flex-child"></div>
    <div class="flex-child"></div>
  </div>

and

.flex-parent {
  display: flex;
  position: absolute;
  top: 0; right: 0; left: 0;

  /*
     IT SHOULD BE POSSIBLE TO SAFELY REMOVE
     THESE TWO LINES BECAUSE THEY ARE OVERRIDEN:
  */
  background: yellow;
  justify-content: center;
}

.flex-parent {
  /* Overriding background: it works! */
  background: green;

  /* Overriding justify-content: NOPE ;-( */
  justify-content: space-between;
}


.flex-child {
  width: 50px;
  height: 50px;
  margin: 10px;
  background: skyblue;
}

Firefox (override works):

enter image description here

Chrome 40 (override seems to be broken):

enter image description here

Am I correct in my assumptions or do I horribly misunderstand CSS? Note that this doesn't have to do with prefixes: Chrome supports unprefixed justify-content, and if it didn't work, there wouldn't be any difference when commenting out rules in first selector.

Update: filed as https://code.google.com/p/chromium/issues/detail?id=451387

Solution

  • You are entirely correct in your understanding of the cascade. If you look in the Web Inspector, you'll see the justify-content: center declaration struck out, but only when you uncheck it (in a similar fashion as commenting it out) can you get Chrome to ignore it.

    Perhaps they accidentally broke something in a change between Chrome 39 and 40, as they have a really irritating habit of doing, but I have no idea what.

    From the issue that rwacarter linked to, apparently they did something funky with their cascade resolution code in order to accommodate certain changes to the Flexbox spec, which seems to be the cause for this. Again, I don't claim to understand their reasons for doing so, but they seem to have a habit of rewriting things here and there that result in regressions all over the place. Good thing Chrome is on an evergreen rapid release cycle, eh?