[Feldt]

What changed / motivation ?

StyleX turns exclusive width bounds into min-width / max-width with a
0.01px offset. Chromium compares min- / max- features with 1/64px
(about 0.0156px) of slack (CompareDoubleValue in media_query_evaluator.cc
uses LayoutUnit::Epsilon()), so these bounds still match at the breakpoint
they are meant to exclude:

• Desktop-first breakpoints (max-width: 900px, then max-width: 600px)
  generate (min-width: 600.01px) and (max-width: 900px). That still matches
  at 600px and wins over (max-width: 600px) there.
• (width > 600px) becomes min-width: 600.01px, which matches at 600px.
• (width < 600px) becomes max-width: 599.99px, which matches at 600px.
• (620px < width <= 900px) from Strict media range loses its excluded boundary in StyleX unplugin/Vite production CSS #1859 becomes
  (min-width: 620.01px) and (max-width: 900px), which matches at 620px. The
  620.01px in the issue comes from StyleX itself (adjustDimension).

#1858 fixed the mirror case (not (min-width: X) -> max-width: X - 0.02px).
This PR moves the offset into one helper, exclusiveBoundValue, used by
adjustDimension (range syntax) and by both negated branches of
mergeIntervalsForAnd, so every exclusive px width bound moves by 0.02px.

The helper has two more rules, so output that was already right stays the same:

• The bound never moves past a whole pixel. A breakpoint written just off a
  pixel keeps landing on it, as on main: not (max-width: 599.99px) and
  (width > 599.99px) give min-width: 600px, and (width < 600.01px) gives
  max-width: 600px. Without this rule, 0.02px would push them to 600.01px
  and 599.99px.
• The result is rounded to 4 decimals, as transform-value.js does for
  numbers, so (width > 0.1px) gives 0.12px, not 0.12000000000000001px.

The whole-pixel rule also moves a few bounds that main pushed past a pixel
back onto it. Mobile-first min-width: 400px then min-width: 600.01px now
gives (min-width: 400px) and (max-width: 600px) instead of 599.99px.
(width > 599.98px) and (width < 600.02px) now give 600px instead of
599.99px and 600.01px.

Heights and non-px units keep 0.01, as in #1858. Chromium has the same slack
for height ((height >= 400.01px) matches in a 400px-tall viewport). I left
height out because #1859 and #1858 are both about width, and changing it
would change more generated class names. Covering px heights too means
dropping the dim === 'width' check and updating the height expectations. I
can add it here if you prefer.

Linked PR/Issues

Fixes #1859

Related: #1763 and #1858 (the same Chromium tolerance, max-width side).

Additional Context

Tests:

• New regression tests: desktop-first breakpoints and the Strict media range loses its excluded boundary in StyleX unplugin/Vite production CSS #1859 query in
  media-query-transform-test.js, plus (width < 600px), (width > 600px)
  and (620px < width <= 900px) in parse-media-query-test.js. On main
  they fail (they get 600.01px, 599.99px and 620.01px).
• New tests for the whole-pixel rule: desktop-first with a 599.99px
  breakpoint, mobile-first with a 600.01px breakpoint, (width > 599.99px)
  and (width < 600.01px). One more, (width > 0.1px), covers the rounding.
  The existing not (max-width: 99.99px) tests are unchanged and still
  expect min-width: 100px.
• Existing expectations for whole-pixel breakpoints move from X.01px to
  X.02px, and from X.99px to X.98px for strict ranges. The class names
  in one babel-plugin snapshot change with them.

Browser check, in local headless Chromium 151 and WebKit 26: I compiled the
styles with the built babel plugin and loaded the CSS at widths 599, 600, 601,
619, 620 and 621. Measured against what the spec gives for the authored
queries, main is wrong at 4 of these points in each engine. With this
change it is wrong at none. matchMedia at 600px in Chromium:
(min-width: 600.015625px) is true and (min-width: 600.0157px) is false,
which is the 1/64px slack. (min-width: 600.01px) and (max-width: 599.99px)
are true, and the 600.02px and 599.98px versions are false.

Breakpoints written just off a pixel (599.99px, 600.01px) are read with
each engine's own slack, so at 600px the engines do not always agree with the
spec. For these, the output in both engines is the same as on main, with
one exception: mobile-first 600.01px at 600px in WebKit. WebKit matches
min-width: 600.01px at 600px, so main showed the 600.01px rule there.
This change shows the 400px rule, which is what the spec gives.

Firefox was not tested (it is not installed here).

Local results on Node 22 with yarn 1.22.22: yarn install --frozen-lockfile
(which builds every package), yarn flow, yarn test:packages, the
devtools-extension build checks, yarn prettier:report and yarn lint:report
pass. The spelling job (typos) was not run locally.

Breaking change: none in the API. The generated CSS for these queries changes
by 0.01px, so their class names change.

Test command: yarn test. For just the parser tests:
cd packages/style-value-parser && npx jest src/at-queries.

This PR was prepared with an AI coding agent (Claude Code). The submitter is
responsible for it and will answer review.

Pre-flight checklist

• I have read the contributing guidelines
  Contribution Guidelines
• Performed a self-review of my code

🤖 Generated with Claude Code

[meta-cla]

Hi @Feldt!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at cla@meta.com. Thanks!

[vercel]

@Feldt is attempting to deploy a commit to the Meta Open Source Team on Vercel.

A member of the Team first needs to authorize it.