The travis job terminated successfully, but after_failure is called

Travis CI Discussions & Feedback
1

This is a problem from yesterday
The SCRIPT part of TRAVIS’s JOB succeeds and runs fine, but I’m having an issue where AFTER_FAILURE is always called.
echo “${TRAVIS_TEST_RESULT}” after each step, but they all return 0 as normal.

$ echo "${TRAVIS_TEST_RESULT}"
0
The command "echo "${TRAVIS_TEST_RESULT}"" exited with 0.
cache.2
store build cache
after_failure.1
0.00s$ echo "${TRAVIS_TEST_RESULT}"
0
after_failure.2
35.14s$ gem install curb
after_failure.3
0.68s$ test/bin/cancel-other-jobs
2

We’re also seeing this same exact issue - it started around 6AM Eastern on March 28, 2023. With no changes to our scripts or travis.yml, all f our jobs returned 0 and ran the after_failure:

The command "INSTALL_MODE="install" ./canary-scripts/deploy.sh" exited with 0.
$ ./canary-scripts/failure.sh

We’ve reached out to Travis Support!

3

We also seeing the same issue started at around 6AM yesterday. Both the after_success.yml and after_failure.yml scripts were run

$ bin/travis/after_success.sh
after_failure
$ bin/travis/after_failure.sh
Done. Your build exited with 0.
4

We experienced the same issue. We had to delete our after_failure step to avoid this unexpected behavior. Are we right thinking that TRAVIS_TEST_RESULT decides if this step fires?

5

Hi everyone, as we’re looking into the issue, here’s the current workaround:

after_success:
  - "[[ $TRAVIS_TEST_RESULT = 0 ]] && echo passed"
after_failure:
  - "[[ $TRAVIS_TEST_RESULT = 1 ]] && echo failed"
6

We’ve encountered this as well. no update after 3 weeks?

7

look at @montana’s reply here on this thread: TRAVIS_BUILD_STAGE_NAME not populated - #5 by Montana

8

Adding here as it seems likely linked. we’re seeing a weird thing where each of the jobs from our stage are being pumped through after_success even though they had failed.

9

Hi @convenient,

We are currently investigating.

1 Like
10

Yeah thanks @Montana , was just adding the comment here for completeness

1 Like
11

(Only found this thread when I went to write my own post).

There’s a worse, possibly connected issue. Deployment steps are being run even when the test (“script”) stage fails.

How long does it take to fix what is a potentially catastrophic problem, if a deployment succeeds after tests failed?

We first noticed it happening around 2024-04-11, though further investigation showed it had happened a couple of days before then. It’s 2024-04-19 and it’s still happening.

It also means that deployment scripts are needlessly being run, just wasting lots of CPU cycles and lots of time.

12

Hi all,

Try this:

deploy:
  ...
  on:
    condition: $TRAVIS_TEST_RESULT = 0

This should have the build pass.

13

We still experience this. Is it a known issue that we must workaround?

14

Hi all,

We are working on a fix for it as we speak.