perf: speed up benchmark checkout - #23211

I'm wondering if that's actually faster. Locally, a git clone https://github.com/php/php-src --filter=blob:none --depth=1 takes twice the time vs no filter.

But what about no depth? That's what the job does, because we need to inspect the history but not the files.

I'm wondering if that's actually faster. Locally, a git clone https://github.com/php/php-src --filter=blob:none --depth=1 takes twice the time vs no filter.

With your --depth=1 is not the same. With it barely any historical blobs have to be filtered/skipped.

But what about no depth? That's what the job does, because we need to inspect the history but not the files.

Exactly. Without --depth=1 the blob:none does not fetch all old file contents but keeps the full commit history, which is what it needs here. My local tests are ~3× faster.

Ok, I got confused by fetch-depth: 0. I confirm that --filter=blob:none without --depth 1 is faster.

Thank you!