fix(fslib): handle float timestamps in convertToBigIntStats

[nelsondude]

Summary

Fixes RangeError when converting floating-point timestamps to BigInt in the convertToBigIntStats function.

Problem

The convertToBigIntStats() function fails when file stats contain non-integer timestamps (e.g., 1763746784088.47), throwing:

RangeError: The number 1763746784088.47 cannot be converted to a BigInt because it is not an integer

This occurs when ZIP file entries have timestamps with floating-point precision, particularly observed in Yarn 6.0.0-rc.5 when building with Storybook.

Solution

Use Math.floor() to ensure integer values before BigInt conversion on line 194 of packages/yarnpkg-fslib/sources/statUtils.ts.

Changes

• Changed BigInt(element) to BigInt(Math.floor(element))

Testing

• Tested with Yarn 6.0.0-rc.5 + Storybook build
• Verified the RangeError no longer occurs with floating-point timestamps

Related

This issue was discovered while running yarn storybook build with Yarn 6.0.0-rc.5.

[arcanis]

One suggestion

[arcanis]

These also need to be updated, as otherwise we lose the millisecond precision. Something like this should work, I think:

bigintStats.atimeNs = bigintStats.atimeMs * BigInt(1e6) + BigInt(Math.floor((stats.atimeMs % 1) * 1e3)) * BigInt(1e3);
bigintStats.mtimeNs = bigintStats.mtimeMs * BigInt(1e6) + BigInt(Math.floor((stats.mtimeMs % 1) * 1e3)) * BigInt(1e3);
bigintStats.ctimeNs = bigintStats.ctimeMs * BigInt(1e6) + BigInt(Math.floor((stats.ctimeMs % 1) * 1e3)) * BigInt(1e3);
bigintStats.birthtimeNs = bigintStats.birthtimeMs * BigInt(1e6) + BigInt(Math.floor((stats.birthtimeMs % 1) * 1e3)) * BigInt(1e3);

[nelsondude]

Updated - great suggestion 👍