Git library in pure Go https://cgit.space/~runxiyu/furgit.git/atom/objectstore/packed/pack_idx_checksum.go?h=master 2026-03-25T14:30:31Z 2026-03-25T14:30:31Z Runxi Yu 2026-03-25T14:30:31Z urn:sha1:bfa0a3f5f18b752a6ebd3d5b37411c6871f7bb17 2026-03-25T14:22:01Z Runxi Yu 2026-03-25T14:21:50Z urn:sha1:e4373a7e553f8523db3e99ea316e1c25d30cc059 2026-03-04T00:59:53Z Runxi Yu 2026-03-04T00:26:56Z urn:sha1:ab7501be34032fb9e5c48726a68ae90a917af9eb 2026-03-03T10:31:11Z Runxi Yu 2026-03-03T10:26:57Z urn:sha1:bd9519b464f919f2ca174a45c3c19c9a8a1fe3d1 We previously had helpers in format/pack/checksum that checks .pack/.idx-related checksums with []byte-based APIs. But it only really makes sense to use those []byte-based APIs on mmap's (otherwise it'd be horribly inefficient). Since the packed object-store only needs to check that the .pack and .idx trailer match, we move the relevant part into objectstore/packed. The rest are deleted for now; we'll definitely need a streaming version for the pack verification (when ingesting packfiles from the network) (though we might just make it a streaming API (writer? reader? not decided yet) that *produces* a hash, then verify it in the caller; this way we could reuse the function in the pack-producing routines). The others might get the []byte-based APIs back, or perhaps they too get streaming APIs. Remember that "reading objects from a packed object store", "creating/writing packfiles", and "ingesting an incoming pack (which usually involves creating an .idx for it)", are all very different tasks.