Conversation
Bumps WebKit to pick up oven-sh/WebKit#354, which adds Karatsuba multiplication, Burnikel-Ziegler division, and a divide-and-conquer toString to JSBigInt. BigInt.prototype.toString, toLocaleString, Intl.NumberFormat.format(bigint), * and ** all inherited the previous O(n^2) behavior; they are now O(n^log2(3)). The new test covers correctness across the 32/64-digit algorithm thresholds and asserts that decimal conversion of a 12500-digit vs 100000-digit value stays within the sub-quadratic growth factor.
|
Warning Review limit reachedYou’ve reached a temporary PR review limit under our Fair Usage Limits Policy. Next review available in: 59 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Run ID: 📒 Files selected for processing (2)
Comment |
|
Updated 1:17 AM PT - Jul 26th, 2026
❌ @autofix-ci[bot], your commit 6d36b91 has 2 failures in
🧪 To try this PR locally: bunx bun-pr 35904That installs a local version of the PR into your bun-35904 --bun |
|
Build currently fails because Verified locally against a I'll push to re-trigger once the WebKit autobuild is up. |
|
Found 2 issues this PR may fix:
🤖 Generated with Claude Code |
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
| * From https://github.com/oven-sh/WebKit releases. | ||
| */ | ||
| export const WEBKIT_VERSION = "549170099226f816a4b204ea1d8fa102fb79eefa"; | ||
| export const WEBKIT_VERSION = "autobuild-preview-pr-354-a1b937f0"; |
There was a problem hiding this comment.
🔴 WEBKIT_VERSION is set to autobuild-preview-pr-354-a1b937f0, a PR-preview release tag rather than a merged main-branch sha. Preview autobuild releases on oven-sh/WebKit are ephemeral and get deleted when the upstream PR closes, at which point prebuiltUrl() will 404 for every build at this commit. As the PR description already notes, this needs updating to the merged main sha before landing.
Extended reasoning...
What the bug is
scripts/build/deps/webkit.ts:6 sets WEBKIT_VERSION = "autobuild-preview-pr-354-a1b937f0", replacing the previous 40-hex commit sha. This is a PR-preview autobuild tag for oven-sh/WebKit#354, not a durable release ref. The file's own doc comment (lines 1-5) says the value is a "WebKit commit — determines prebuilt download URL + what to checkout for local mode … From https://github.com/oven-sh/WebKit releases", and the PR description explicitly states: "WEBKIT_VERSION is currently pointed at the PR-preview release for oven-sh/WebKit#354; it needs updating to the merged main sha before this lands."
Code path that triggers it
In prebuilt mode (the default for CI and most contributors), webkit.source(cfg) returns { kind: "prebuilt", url: prebuiltUrl(cfg), … }. prebuiltUrl() computes:
const tag = version.startsWith("autobuild-") ? version : `autobuild-${version}`;
return `https://github.com/oven-sh/WebKit/releases/download/${tag}/${name}.tar.gz`;So the download URL becomes …/releases/download/autobuild-preview-pr-354-a1b937f0/bun-webkit-<os>-<arch><suffix>.tar.gz. That release exists right now because the WebKit PR is open, but preview autobuild releases are cleaned up once the PR merges/closes. After cleanup, this URL 404s and every fresh build (bun bd, CI, release builds) fails at the WebKit fetch step.
Why existing code doesn't prevent it
The build system does have first-class handling for autobuild- tags — prebuiltUrl() skips re-prefixing, and prebuiltDestDir() uses the whole tag as the cache-dir key rather than slice(0, 16). That handling exists precisely so preview tags can be tested via --webkit-version=<tag> (per the doc comment) without corrupting the cache. It makes the preview tag work today; it does nothing to make the ref durable. There is no fallback URL, no pinned mirror, and local mode (vendor/WebKit/) checks out the same value, which is not a valid commit sha to git checkout.
Impact
Landing as-is pins main to an artifact with a lifetime measured in days. Once oven-sh/WebKit#354 merges and its preview release is deleted:
- Every CI build on any branch based on this commit fails at dependency fetch.
- Every contributor without a cached
webkit-preview-pr-354-a1b937f0*extraction cannot build. git bisectacross this commit becomes impossible without manual WebKit surgery.
This is a concrete regression in something that worked before (builds at any historical commit can fetch their pinned WebKit), which is why it's marked normal rather than nit despite the author already being aware — flagging it here prevents an accidental merge before the sha swap.
Step-by-step proof
cfg.webkit === "prebuilt"(default) →source()returnsprebuiltUrl(cfg).cfg.webkitVersion = "autobuild-preview-pr-354-a1b937f0";version.startsWith("autobuild-")is true, sotag = versionverbatim.- Fetch URL:
https://github.com/oven-sh/WebKit/releases/download/autobuild-preview-pr-354-a1b937f0/bun-webkit-linux-amd64.tar.gz. - JSBigInt: sub-quadratic multiply, divide, and toString WebKit#354 merges → its
autobuild-preview-pr-354-*release is garbage-collected per the WebKit repo's preview-release policy. - GitHub returns 404 for the URL in step 3 → build aborts before compilation.
Fix
After oven-sh/WebKit#354 merges, replace the value with the resulting 40-hex main-branch commit sha (the one that carries the Karatsuba/Burnikel-Ziegler/D&C-toString changes), matching the format of the previous value 549170099226f816a4b204ea1d8fa102fb79eefa. No other code changes needed — prebuiltUrl() and prebuiltDestDir() already handle plain shas correctly.
There was a problem hiding this comment.
Acknowledged; this is intentional while oven-sh/WebKit#354 is open so CI builds against the change. It will be swapped to the merged main sha before this lands (also noted in the PR description).
|
Closing in favor of #40299. This PR bumps WebKit to oven-sh/WebKit#354. That PR is superseded by oven-sh/WebKit#507, which is merged. #507 ports V8's sub-quadratic multiply, divide, #40299 is the WebKit bump for #507 and fixes #39964. Its |
2 participants
BigInt.prototype.toString(and everything that uses it:toLocaleString,Intl.NumberFormat.prototype.format(bigint), template literals) is quadratic in the digit count. So are BigInt*and**. Every doubling of the digit count quadruples the time:Cause
JavaScriptCore's
JSBigInt::toStringGenericrepeatedly divides by a one-digit chunk divisor (10^19on 64-bit), so an n-digit value takes ~ndivideSinglepasses of ~n work each. The only multi-digit multiply is schoolbook, so*and**are quadratic too.Fix
oven-sh/WebKit#354 adds Karatsuba multiplication (used by
*,**), Burnikel-Ziegler recursive division, and a divide-and-conquertoStringthat splits by repeatedly-squared powers of the chunk divisor. All three paths are O(n^log2(3)) instead of O(n^2).This PR bumps
WEBKIT_VERSIONto that change and addstest/js/bun/jsc/bigint-subquadratic.test.ts, which checks correctness across the 32/64-digit algorithm thresholds and asserts that the 12500-to-100000-digittoStringgrowth stays within the sub-quadratic factor.Measured (Linux x64 release, best of three; node v26.3.0 for reference)
toString(10)toString(10)toString(10)Scaling per doubling drops from x4.0 to x2.9. Decimal
toStringnow beats node. Multiply is 7.5x faster but V8 is still ahead (it also has Toom-3 and FFT above Karatsuba).Note:
WEBKIT_VERSIONis currently pointed at the PR-preview release for oven-sh/WebKit#354; it needs updating to the merged main sha before this lands.