[Bug]: Issue with unitless zero, starting with 11.2.0

[bazineta]

Describe the bug

Deep in the bowels of bootstrap's CSS, there's code that produces this:

padding:calc(.375rem - 1px - 0) .75rem calc(.375rem - 1px - 3px - 0)

The 11.1.2 release does not complain about this, but starting in 11.2.0, and continuing in 11.2.1, the following warning is issued:

Invalid CSS calculation type

I suspect it's not dealing with the unitless zero in the calculation in the same way it did previously.

Expected behaviour

I expect this to not issue a warning, and instead to determine that zero is....zero, and therefore uninteresting.

Steps to reproduce

Using this minimal repro:

.x {
  padding: calc(.375rem - 1px - 0);
}

Will, under 11.1.2, produce:

x {
  padding: calc(.375rem - 1px);
}

Using 11.2.0 or 11.2.1, it will instead produce:

Invalid CSS calculation type

Version

11.2.1

Environment

System:
    OS: macOS 26.7
    CPU: (16) arm64 Apple M3 Max
    Memory: 8.04 GB / 64.00 GB
    Shell: 5.9 - /bin/zsh

Package details

├─┬ autoprefixer@10.6.1
│ └── postcss@8.5.28 deduped
├─┬ css-loader@7.1.5
│ ├─┬ icss-utils@5.1.0
│ │ └── postcss@8.5.28 deduped
│ ├─┬ postcss-modules-extract-imports@3.1.0
│ │ └── postcss@8.5.28 deduped
│ ├─┬ postcss-modules-local-by-default@4.2.0
│ │ └── postcss@8.5.28 deduped
│ ├─┬ postcss-modules-scope@3.2.1
│ │ └── postcss@8.5.28 deduped
│ ├─┬ postcss-modules-values@4.0.0
│ │ └── postcss@8.5.28 deduped
│ └── postcss@8.5.28 deduped
├─┬ cssnano@9.0.5
│ ├─┬ cssnano-preset-default@9.0.5
│ │ ├─┬ cssnano-utils@7.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-calc@11.2.1
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-colormin@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-convert-values@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-discard-comments@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-discard-duplicates@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-discard-empty@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-discard-overridden@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-merge-longhand@9.0.4
│ │ │ ├── postcss@8.5.28 deduped
│ │ │ └─┬ stylehacks@9.0.4
│ │ │   └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-merge-rules@9.0.4
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-minify-font-values@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-minify-gradients@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-minify-params@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-minify-selectors@9.0.4
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-normalize-charset@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-normalize-display-values@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-normalize-positions@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-normalize-repeat-style@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-normalize-string@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-normalize-timing-functions@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-normalize-unicode@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-normalize-url@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-normalize-whitespace@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-ordered-values@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-reduce-initial@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-reduce-transforms@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-svgo@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ ├─┬ postcss-unique-selectors@9.0.3
│ │ │ └── postcss@8.5.28 deduped
│ │ └── postcss@8.5.28 deduped
│ └── postcss@8.5.28 deduped
├─┬ postcss-loader@8.2.1
│ └── postcss@8.5.28 deduped
├── postcss@8.5.28
└─┬ resolve-url-loader@5.0.0
  └── postcss@8.5.28 deduped

Additional context

No response

[ludofischer]

What version of Bootstrap is this? According to the CSS specification, that expression should be invalid:

That is, width: calc(0 + 5px); is invalid, because it’s trying to add a to a , even though both width: 0; and width: 5px; are valid.

https://drafts.csswg.org/css-values/#calc-type-checking

Is it possible that expression is created by a previous transformation step? Why a stylesheet author add 0?
I Firefox at least, the calc() declaration is ignored if you add 0 to a unit.

[bazineta]

Finally traced to the tom-select plugin, rather than bootstrap itself; admittedly the transformation pipeline makes finding the culprit a bit tedious. They had indeed used a unitless zero there, and correcting it to 0px eliminated the warning without any other regressions. This can be closed, but I'd be surprised if others do not trip over similar issues, so perhaps this report will be useful in that regard.