Sounds reasonable to me I suppose.

I think pointer-events: none is here so that you cannot click on the button when it is disabled. If you remove this style, disabled buttons will look like they are disabled but will be in fact enabled (you can click on them and trigger an event).

@steph643 We basically don't recommend using the .disabled class on <button>s if possible; use the disabled attribute instead (which itself prevents clicks). <a> doesn't support the disabled attribute (at least not yet), so we use pointer-events: none there to imperfectly approximate the prevention of clicks.

@cvrebert, thanks a million. I had this mistake in several places in my code.

In our local builds I've chosen to add this rule back until we've been able to audit our existing codebase as we've been relying on the effect of .disabled across all .btn-able elements.

I understand the reason for this specific issue, but this type of breaking change seems awfully severe for a bugfix release—whereas .btn.disabled worked consistently before in actually disabling click events across all .btn elements with or without the disabled attribute, this means updating existing code to also set the disabled attribute on buttons.

Given the benefit of this backwards-incompatibility seems limited to styling on mouse hover events, I'd argue that this change would be suitable for a major release (4.0) and not for a bugfix or even minor version bump where compatibility is to be preserved.

FWIW, our docs only show/mention usage of .disabled on <a>s. I checked pretty-old v3.0.2 and couldn't find any instances of class="disabled" on <button>s in those docs either.

I will see about adding a linter for button.btn.disabled to Bootlint to help folks avoid this.
It might also be worth adding an explicit warning about this to the docs.

Is it safe to override the pointer-events: none with

 .btn[disabled] {pointer-events: auto;}

or will I get unexpected behavior? I rather like having the cursor: not-allowed on my disabled buttons.

It's a breaking change! Why was it merged to v3?? Now I have to rewrite some code :(
Thanks @cvrebert for explanation how to fix.