Travis not running on a clean checkout (was Re: Can any one explain the Travis/CI errors on PR #7639 : LayoutEditor Code Coverage and Test Cases

Bob Jacobsen

Again, there are two issues here:

1) The script was flawed; it was using ant as part of its process when a file was not present. That’s being fixed in PR 7658 to not use ant any more (it explicitly invokes ant via maven instead)

2) Somehow, Travis jobs are running _with_ prior-built content still in place. This caused (1) to not be detected until now, because the file _was_ (mostly) still present. That’s a more-troubling problem


On Nov 24, 2019, at 8:19 AM, Randall Wood via Groups.Io <> wrote:

The latest build from master and the build that failed in the PR are from different Travis CI build server recipes; one of which includes ant and one that does not. I can’t find anything in that suggests that ant is a standard part of the Travis CI, and we may have lucked out with it being present. suggests that ant is *not* a standard part of Travis CI, and includes instructions on how to make it available (note this takes up build time).

On 24-Nov-2019, at 10:40, Bob Jacobsen <@BobJacobsen> wrote:

The underlying cause is a problem in the Travis job `scripts/` script. The branch for the 4th Travis job runs (only) the `scripts/run_flagged_tests_separately` script, which in turn uses `./runtest.csh` to run specific tests.

This has worked so far, but it assumes that (1) the code has been compiled and (2) the hook has been created by Ant. This branch of the `` script doesn’t specifically do either of those. I’ll fix that this morning.

The more interesting things is that this has _worked_ most of the time so far. Where did the class files and come from then? Some prior build?


On Nov 24, 2019, at 5:03 AM, danielb987 <db123@...> wrote:

George Warner wrote that it was PR #7639. So the logs are here:
The error is on Travis HEADLESS=false SKIPINTERMITTENT=false STATIC=true


2019-11-24 13:55 skrev Randall Wood via Groups.Io:
When being asked to look at the logs of a failing build, we need to
have the complete logs to look at, since the real failure may be
None of our CI environments support ant, so (at least on AppVeyor) we
use maven to stage ant, but I can’t tell what failed here since any
logs about how ant was being installed/configured are not in the
snippet you show.

Bob Jacobsen

Bob Jacobsen