Closed rocky closed 4 years ago
@boretom this may be one of the reasons you are seeing failures on some systems.
On the systems this dash-lowercase-c
works on, the dash-l
test starts after dash-lowercase-c
finishes. On systems where this fails the directory is set incorrectly presumably because dash-l
hasn't finished and has the changed the current working.
Change what I wrote before.
Failure is of dash-l
has to do with where make is looking for the file it is testing.
On a working system, it looks in remake/tests/work/options
. On a failing system, when I cd to that directory in the makefile right before the test inside the Makefile, things work.
What's weird is that I can take a binary from one Ubuntu system where this works, copy it to a system where it fails, and the code will still fail. So there is something about the OS or environment, rather than a way the code has been compiled that causes a problem.
I'm closing this PR, but will leave the branch around to experiment with should you wish to do so.
Canging directories is having a bad effect on surrounding tests, e.g. dash-l which sometimes can get run in parallel with dash-lowercase-c
The test framework really isn't set up for making directories and removing them cleanly afterwards, or isolating each tests environment so one can't have a bad effect on another.
And the test code here isn't doing the trick either.