oao yarn invocation errors often give insufficient information about oao HOT 4 CLOSED

Hmmmm, I'm afraid oao is not generating the Killed message. It is indeed the stderr output: in the following line, notice the yarn bit (source of the message: either stdout or stderr of the command being launched) and the ERROR bit (hence, stderr):

2017-10-10T14:10:03.867Z                 yarn ERROR | Killed

Wrt. other comments:

• The missing yarn header: maybe yarn gets killed (by whom?) before printing the header out?
• I agree we should print out the exit code.
• I don't understand your 4th bullet. In principle, it does run yarn link neutrino from the packages/neutrino-middleware-env directory (in the first bootstrap phase, it had already run yarn link from the packages/neutrino directory). Maybe I'm missing something?
• I agree with the unnecessary duplication, but it would have a lower priority right now…

from oao.

Ah perhaps it's an OOM?
https://stackoverflow.com/questions/10876136/what-can-cause-node-js-to-print-killed-and-exit

I believe we could confirm that by seeing if the exit code was 137 (http://www.tldp.org/LDP/abs/html/exitcodes.html).

The missing yarn header: maybe yarn gets killed (by whom?) before printing the header out?

If it's an OOM I would have thought it would occur after the point where yarn has output the version header. As such I would guess either:

• the output is being buffered (which is not necessarily a bad thing, but perhaps the buffering can be made to flush more often?)
• execa (or oao) is ignoring the stdout when there's something on stderr to display (I would find this unlikely)

I don't understand your 4th bullet. In principle, it does run yarn link neutrino from the packages/neutrino-middleware-env directory (in the first bootstrap phase, it had already run yarn link from the packages/neutrino directory). Maybe I'm missing something?

Ah I was thinking of the first stage, I don't fully understand what oao does after that.

I agree with the unnecessary duplication, but it would have a lower priority right now…

Agreed :-)

from oao.

Patched in v0.10.4 to log error code upon command failure.

I couldn't find traces of buffered stderr output. In fact, in my examples I got stdout and stderr interspersed, as could be expected in some cases. Example:

2017-10-10T20:45:06.934Z                 main INFO  ┌── Run cmd yarn add dfjdfdifdjfdijfdijfdoifjdiojfoi at packages/neutrino [CREATED]
2017-10-10T20:45:07.243Z                 yarn INFO  | yarn add v1.1.0
2017-10-10T20:45:07.384Z                 yarn INFO  | [1/4] Resolving packages...
2017-10-10T20:45:09.441Z                 yarn ERROR | error Couldn't find package "dfjdfdifdjfdijfdijfdoifjdiojfoi" on the "npm" registry.
2017-10-10T20:45:09.441Z                 yarn INFO  | info Visit https://yarnpkg.com/en/docs/cli/add for documentation about this command.
2017-10-10T20:45:09.451Z                 yarn INFO  |
2017-10-10T20:45:09.454Z                 main ERROR Command 'yarn add dfjdfdifdjfdijfdijfdoifjdiojfoi' failed [1] at packages/neutrino
2017-10-10T20:45:09.456Z                 main INFO  └── Run cmd yarn add dfjdfdifdjfdijfdijfdoifjdiojfoi at packages/neutrino [CLOSED]
2017-10-10T20:45:09.458Z           storyboard INFO  └── ROOT STORY: Node.js 8.6.0 on Darwin 64-bit, SB 3.1.3 [CLOSED]

from oao.

Many thanks :-)

from oao.