go-ipfs version: 0.5.0-dev-49afb3b5b-dirty
Repo version: 7
System version: amd64/linux
Golang version: go1.14
Steps to reproduce:
$ yes "foo" | head -n 40000 > test
$ ipfs add test --chunker=buzhash --pin=false
added QmX4kkx8AWgy6GmF9zVtgPVta5GtpEJpPTtatexZ8xrwBs test
156.25 KiB / 156.25 KiB [============================================================================================================================================================================] 100.00%
$ ipfs dag get QmX4kkx8AWgy6GmF9zVtgPVta5GtpEJpPTtatexZ8xrwBs
{"data":"CAIYgOIJIIDiCSAA","links":[{"Cid":{"/":"QmcNJ4CKYGuya2ndNr9ymV7oQsT4S69SdwmTzaiW7c9vAC"},"Name":"","Size":160014},{"Cid":{"/":"QmaRwA91m9Rdfaq9u3FH1fdMVxw1wFPjKL38czkWMxh3KB"},"Name":"","Size":8}]}
$ ipfs add test --chunker=buzhash --pin=false --raw-leaves
added QmYyFBARF9M2bwfBQ7y5kNZGJVJWmktpx2tapZy6gWoBXC test
156.25 KiB / 156.25 KiB [============================================================================================================================================================================] 100.00%/
$ ipfs dag get QmYyFBARF9M2bwfBQ7y5kNZGJVJWmktpx2tapZy6gWoBXC
{"data":"CAIYgOIJIIDiCSAA","links":[{"Cid":{"/":"bafkreigqq7maj23te3dkn4j76n7haab7lgmdz5fztubrdgsv44k4vyipsa"},"Name":"","Size":160000},{"Cid":{"/":"bafkreihdwdcefgh4dqkjv67uzcmw7ojee6xedzdetojuzjevtenxquvyku"},"Name":"","Size":0}]}
Adding a file with --chunker=buzhash --raw-leaves results in two blocks, whose sizes are 160000 and 0 respectively.
EDIT: The trailing block with non-raw-leaves add contains no actual data too, so this is not related to raw-leaves.
Definitely a bug. I will try to dig into it before next Monday ( reimplementing the buzzhash convergence in another tool is something I was going to work on tomorrow )
@ribasushi status?
@Stebalien I haven't dug into this yet: it can't go into 0.5 anyway as it will be changing hashes. Or do we actually want this for 0.5 ( in which case I will speed it up ).
Well, (a) this is a bug, (b) this feature is experimental, and (c) it has never been released. So I'd like to fix it before the release and before users start to actually use it.
(c) it has never been released
I did not realize this, apologies. Fix will be here by tomorrow.
Most helpful comment
Definitely a bug. I will try to dig into it before next Monday ( reimplementing the buzzhash convergence in another tool is something I was going to work on tomorrow )