test-autopatch is currently very noisy when running ninja check.
This patch redirects much of that noisy output to a log file while still allowing
errors to output to the terminal during ninja check.
Details
- Reviewers
deadalnix - Group Reviewers
Restricted Project
ninja check-arcanist # output shows PASSED without unnecessary info. verify log file in <build>/contrib/arcanist/
Diff Detail
- Repository
- rABC Bitcoin ABC
- Branch
- ap-test
- Lint
Lint Passed - Unit
No Test Coverage - Build Status
Buildable 8393 Build 14804: Default Diff Build & Tests Build 14803: arc lint + arc unit
Event Timeline
contrib/arcanist/test/test-autopatch.sh | ||
---|---|---|
8 | Reviewer note: I removed set -x here because it outputs to stderr by default and isn't strictly necessary. Under some circumstances, it's possible to change that to stdout by setting export BASH_XTRACEFD=1 but this appears to affect subshells in strange ways. |
That whole thing depends on a gazillion environments variables, internet connectivity, and whatever else, and doesn't check at all that bitcoin-abc is working. I really don't see why someone who want to checkout the ABC sources and run the tests to make sure it works have to go through the bullshit of installing arc, setting up an auth token and alike. That makes no sense whatsoever. The software is supposed to get better *FOR USERS* over time. What does it do in the middle of ninja check? How come, when the output of ninja check is polluted, nobody stops one second to ask themsleves "Wait, why is that thing in the output to begin with? Does it need to be?" but instead go on changing all the output formatting so that the main problem remains, but is less visible?