Travis doesn't catch failing installations

[brucellino]

In Travis, we run the installation of a product in a container - it recently failed (see #17 ) silently. Since the installation is done in the container, Travis sees an exit 0 of docker exec, not the exit 1 of rpm -ivh

Somewhere between the python script, the rpm post we are losing the error code.

[gwarf]

@brucellino, may we try something like:
docker/compose#3379 (comment)

[gwarf]

It seems it doesn't work with actual .travis.yml, need to dig more why error code is being lost.

[gwarf]

So it seems that in fact the problem is that a failing postscript does not make the install fail!

[root@beeb16c596bf src]# rpm -Uvh build/RPMS/noarch/glite-info-update-endpoints-3.0.1-1.el7.noarch.rpm  ; echo $?
Preparing...                          ################################# [100%]
Updating / installing...
   1:glite-info-update-endpoints-3.0.1################################# [100%]
warning: %post(glite-info-update-endpoints-3.0.1-1.el7.noarch) scriptlet failed, exit status 1
0

[gwarf]

So it's a known thing: https://bugzilla.redhat.com/show_bug.cgi?id=569930

[gwarf]

Move away from Travis, not planning to hack a workaround for this right now.