t0080 improvements

[chriscool]

This may not fix #2028.
But it might help us get more information when it fails.

[chriscool]

Here is the (unrelated) Travis-CI failure in t0230:

expecting success: 

        mkdir -p testdir &&

        echo "hello test" >testdir/test.txt &&

        ipfs add -r testdir &&

        curl -i "http://localhost:$PORT_API/api/v0/refs?arg=QmTcJAn3JP8ZMAKS6WS75q8sbTyojWKbxcUHgLYGWur4Ym&stream-channels=true&encoding=text" >actual_output



No output has been received in the last 10m0s, this potentially indicates a stalled build or something wrong with the build itself.

The build has been terminated

[chriscool]

Circleci build also timed out but in t0040!

[rht]

@chriscool t0080, t0043, t0081, t0160 all contain the tail -n1 race condition. Perhaps "ipfs add x then get the last hash" can be wrapped in a function.

[jbenet]

Perhaps "ipfs add x then get the last hash" can be wrapped in a function.

yes, sounds like a good idea. we could also make this into another output mode for ipfs add, which only outputs the root hash. (it's a common enough occurrence that forcing | tail -n1 always is annoying). We could make -q do this, or use -qq*

• @whyrusleeping another reason why -qq is not great for --silent. there may be degrees of quiet before silent

[jbenet]

I'm sick of these osx travis failures. we need to find the problem and fix it.

[jbenet]

@rht, wait

all contain the tail -n1 race condition.

whats the race condition?

[chriscool]

@jbenet

we could also make this into another output mode for ipfs add, which only outputs the root hash. (it's a common enough occurrence that forcing | tail -n1 always is annoying).

Yeah, I think it's a good idea!

@rht I don't like ipfs cmd ... | something because it doesn't detect when ipfs cmd ... fails as by default the exit code of the whole expression is the exit code of something, the part after the pipe (|). (You can try false | true; echo $? in a shell to see that.)

[rht]

@chriscool / @jbenet I see, I thought the cause of the fluctuating ipfs refs -r in https://github.com/ipfs/go-ipfs/pull/2041/files#diff-3300d84a935a8de682f5e1225c53b384L220 was due to a tail -n1 race condition, and making sure ipfs add -r -q . complete first was to avoid this.

[chriscool]

Is there something that should be done on this?
The build failure are not related.

[jbenet]

thanks @chriscool

[rht]

Found several errs of this type while stress testing in #2101, e.g. t0045 https://travis-ci.org/ipfs/go-ipfs/jobs/97678447 and one more in t0080 refs -r(#2028). At least in the t0045 case it is shown that some of the files don't get added.

[chriscool]

Yeah, my commits in this PR haven't fixed the problem described in #2028. It really looks like a bug in ipfs add.