Integrate postcss-flexbugs-fixes into build

[cvrebert]

Fixes #18569.
See https://github.com/luisrudge/postcss-flexbugs-fixes
@mdo SGTY?

[mdo]

Seems fine to me. We'll want to document that IE bug a bit more though no? Is there no way to keep a hack inline for it?

[cvrebert]

Is there no way to keep a hack inline for it?

@mdo Not that I'm aware of. The alternative is to be constantly vigilant when writing our CSS to:

• Never use the flex shorthand with less than 2 values
• Never use unitless 0 as the 3rd value in flex shorthand; always use 0% instead

(Or always use the longhand properties instead of the flex shorthand property.)

[cvrebert]

Or you could declare IE10-11 unsupported for Flexbox. They have many Flexbox bugs; see https://github.com/philipwalton/flexbugs/blob/master/README.md

[mdo]

Hmm, we could to the first bullet you mention, but not the second (unless we disable our linter for unitless values). Maybe we try that instead to avoid an extra dependency?

[cvrebert]

I suppose we could ask/PR SCSS-Lint to (1) add a new rule for the 1st bullet and (2) add the ability to exempt certain properties from the ZeroUnit rule. But that's 2 birds in the bush versus a bird in the hand.

[slavanga]

Please also keep in mind that a lot of people compile Bootstrap with apps like Codekit or Prepros. They won't be able to use postcss-flexbugs-fixes.

[cvrebert]

I note that the same argument could have been made about Autoprefixer a while ago. Being chained to slow-moving GUI tools in the hyper-fast-paced web frontend ecosystem doesn't strike me as the wisest decision.

[cvrebert]

Filed sds/scss-lint#702 and sds/scss-lint#703
So now we're at the mercy of someone being generous enough to implement those, if we go the SCSS-Lint route.

[cvrebert]

Going with this for now.