unixfs: refactor precalcNextBuf

[schomatis]

Addresses part of #5192.

---

unixfs: refactor switch in `precalcNextBuf`

Do not use `NewDagReader` just for the `RawNode` case.
Treat invalid UnixFS types in the same case.

---

unixfs: split `precalcNextBuf`

Create new `loadBufNode` function to handle the `buf` logic which is unrelated
to the main `precalcNextBuf` logic of processing promises to fetch nodes.

[Stebalien]

Needs rebase but otherwise LGTM

[schomatis]

Rebased, but there are some weird Travis errors I should look into.

[Stebalien]

Those errors look like #5070

[schomatis]

Yes, I'm used to Jenkins failing (I don't really pay attention to it anymore), but I've never seen Travis fail like that (I mean passing only one of the sharness tests and one of the Go tests from different architectures). I'm leaving the link here to document it and I'm making push to trigger another test.

https://travis-ci.org/ipfs/go-ipfs/builds/404523422?utm_source=github_status&utm_medium=notification

• OSX: /Users/travis/.travis/job_stages: line 187: shell_session_update: command not found (may be related to shell_session_update error travis-ci/travis-ci#6307)

[schomatis]

And the Linux error is on me :)

(edit: thought I was causing the Linux error in test/sharness/t0087-repo-robust-gc.sh but seems to be another spurious error)

[Stebalien]

Yeah, I get the travis error all the time. We really need to fix that test.

[whyrusleeping]

@schomatis youre on fire, I think i've merged more PRs from you in the past few days than i've merged for myself all year o.o

[schomatis]

Thanks, they are mostly small and cosmetic changes, but I prefer to have (many) compact PRs (instead of a single massive one) because they tend to be easier to review and reason about. (On the other hand they open the door to some unexpected conflicts like #5239 :))