Add flush command to ipfs files

[whyrusleeping]

No description provided.

[jbenet]

this notation is a bit hard to read and error prone. i think these may be better:

nd.Blockstore.PinLk.Lock()
defer nd.Blockstore.PinLk.Unlock()
// or
defer nd.Blockstore.PinLock().Unlock()

[jbenet]

@whyrusleeping this looks good. is it safe ?

[whyrusleeping]

@jbenet its too safe. I need to remove the pinning.

[jbenet]

its too safe. I need to remove the pinning.

ahhaahah

[jbenet]

[whyrusleeping]

alright, this is ready. most of the review is needed on the last commit, i ripped the mfs file apart into an 'inode' type object, and a 'file descriptor' type object. This makes working with file things much easier, and i've been meaning to do it for a long time. Things should be a little more performant, way more threadsafe, the files api stuff shouldnt hang for @diasdavid, and fuse should be much happier in general.

[Kubuxu]

From command info I can see that it causes changes not to be flushed to disk. Wouldn't it be better to call it --no-flush?

[whyrusleeping]

@Kubuxu the default is that changes are flushed to disk. the existing --flush command defaults (when not specified) to true. If you want to not flush changes to disk, you specify --flush=false

[Kubuxu]

Ahh, ok. Sorry for that.

[whyrusleeping]

@Kubuxu no worries, must be a documentation issue, what can i change to make that more clear?

[Kubuxu]

Just this wording is unfortunate. It suggests that just --flush is unsafe.

[whyrusleeping]

@Kubuxu there, that seems a little less confusing to me

[whyrusleeping]

@RichardLitt adding 'ipfs files flush' in this PR.

[RichardLitt]

Thanks @whyrusleeping. Noted in ipfs-inactive/http-api-spec#57