Use newer js-ipfs proxy libraries #288
Conversation
This might be due to js-ipfs's change to a new CID library. They've now moved on to the |
I think you ment to say "doesn't garbage collect", right? |
Nah it's fine, we'll power through until your
Yeah, for me pinning is synonymous to "don't garbage collect". |
icidasset
force-pushed
the
icidasset/no-pin
branch
from
169bdfb
to
8756d33
Compare
| @@ -17,5 +17,5 @@ export const setup = { | |||
| user: "fission.name" | |||
| }, | |||
|
|
|||
| shouldPin: true, | |||
| shouldPin: false, | |||
There was a problem hiding this comment.
I think we can just remove this now, right?
I'm not 100% sure. There might be a use-case for pinning for users who're providing their own IPFS instance.
... But at the same time I think there's noone who's actually doing that. 😄
There was a problem hiding this comment.
Can't hurt to keep it around I guess? In case js-ipfs does turn on GC, like how it's supposed to work 😅
icidasset
force-pushed
the
icidasset/no-pin
branch
from
1832b3f
to
0686984
Compare
| return new dagPb.DAGLink(raw.Name, raw.Tsize, raw.Hash) | ||
| const h = raw.Hash | ||
| const c = new OldCID(h.version, h.code, h.multihash.bytes) | ||
| return new dagPb.DAGLink(raw.Name, raw.Tsize, c) |
There was a problem hiding this comment.
This is a bit dumb, but I didn't want to put too much work in here, since this is getting replaced soon anyway by WNFS v2.
| fs.mkdirSync(dir) | ||
|
|
||
| const memoryDs = new MemoryDatastore() | ||
| const memoryBs = new MemoryBlockstore() |
There was a problem hiding this comment.
@matheus23 Bit of double effort here, but needed to replace this to get the tests working with the new ipfs-core.
|
Ready for review. Goes hand in hand with this lobby PR: fission-codes/auth-lobby#92 Which I'll need to deploy to staging to get all the tests passing. |
tests/helpers/loadCAR.ts
Outdated
| @@ -13,7 +13,7 @@ export async function loadCAR(filepath: string, ipfs: IPFS): Promise<{ roots: CI | |||
| const blockIterator = await CarBlockIterator.fromIterable(inStream) | |||
| for await (const block of blockIterator) { | |||
| cids.push(block.cid) | |||
| await ipfs.block.put(block.bytes, { cid: block.cid }) | |||
| await ipfs.block.put(block.bytes, { version: 1, mhtype: "sha2-256" }) | |||
There was a problem hiding this comment.
I think this code has issues, because it'll strip away the IPLD codec from all CIDs and turn all blocks to "raw" blocks implicitly (this code is probably from the wnfs2 branch).
We should totally get rid of this whole thing and just use the newly implemented ipfs.dag.import api.
There was a problem hiding this comment.
Aight! I'll see if I can use that import function 👀
|
The tests are failing because, ... uh. They depend on staging infrastructure lol. |
Gets rid of our
ipfs-message-port-*forks and versions the ipfs worker, because said libraries aren't backwards compatible (due toCIDimplementation changes).How the versioning works:
ipfs.htmlpath, so older webnative clients know where to look.ipfs/v2.htmlwhich works as before, but with the new message-port libraries.Will update to stable version which comes out soon.
Closes #250