Setting "deep" config options doesn't work

[seidtgeist]

I consulted the new CORS documentation in ipfs daemon --help and found this

You can setup CORS headers the same way:

ipfs config --json API.HTTPHeaders.Access-Control-Allow-Origin '["*"]'
ipfs config --json API.HTTPHeaders.Access-Control-Allow-Methods '["PUT", "GET", "POST"]'
ipfs config --json API.HTTPHeaders.Access-Control-Allow-Credentials '["true"]'

However, when I tried it I got an error at first.

Using ipfs config show I saw that API.HTTPHeaders is set to null, so I set it to an empty object like this:

ipfs config --json API.HTTPHeaders '{}'

Afterwards I was also able set properties within the API.HTTPHeaders object.

I would have expected the ipfs config --json command to automatically create objects along a valid property path.

[jbenet]

I would have expected the ipfs config --json command to automatically create objects along a valid property path.

@ehd yeah, it should. interesting... can you repro with a test case? we should add it to test suite if so

[seidtgeist]

@jbenet Yeah, gladly. Should I add it in this test?

I also found this test in user-config which sets a deep property successfully. Lo and behold, maybe a difference between null and not defined in the code that checks deep property access?

One more question: Where do I put the ipfs binary in order to run the tests? From the README I couldn't find out how to build ipfs in a way that makes it work within the test harness command from .travis.yml.

[jbenet]

@ehd yeah that test works.

Where do I put the ipfs binary in order to run the tests? From the README I couldn't find out how to build ipfs in a way that makes it work within the test harness command from .travis.yml.

just run make. it should build its own version and place it in test/bin/ipfs

[seidtgeist]

@jbenet seidtgeist@c6d1413 adds test cases for:

• config set a path with a parent that is not defined: Passes
• config set a path with a parent that is null (like setting a property inside API.HTTPHeaders & Gateway.HTTPHeaders): Fails

[jbenet]

thanks @ehd!

[jbenet]

@rht, @whyrusleeping, @Heems can one of you pick this up this week?

[rht]

Waiting for merge on @ehd's commit before this is rebased and merged: #1570

[seidtgeist]

thanks! should i open a PR or something to get the above commit merged?

[rht]

@ehd uh huh

[seidtgeist]

@rht I opened a PR pulling my commit into your fork so master doesn't get a failing test merged in. Please let me know if you want me to do something else instead.

[seidtgeist]

high 🙏🏻 five. Thank you.