v2.7.4
Cache restores no longer stall on huge lockfiles
A restore that hashed a lockfile over 40 MB could hang for minutes. It now streams the hash.
Some monorepos produce lockfiles larger than 40 MB. Cinder hashed them in memory before restoring a cache, which made the first step of a run look frozen.
What we fixed
The lockfile hash is now streamed, so memory use stays flat and the restore starts at once.
A cache miss caused by a trailing newline change is now reported in the log, with both hashes.
Good to know
Existing caches stay valid. The hash result is identical, only the way it is computed changed.
Get the patch notes.
One short email for each release. Unsubscribe in one click.
Partner link. This demo form saves to your Framer inbox; no emails are sent from this template.