stack.sh does not set -e, so many error cases can go unhandled
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
devstack |
Fix Released
|
Low
|
Unassigned |
Bug Description
e.g. from the run where I noticed that this was the case:
+ timeout 60 sh -c 'while ! http_proxy= curl -s http://
+ SERVICE_ENDPOINT=http://
+ ADMIN_PASSWORD=
+ SERVICE_
+ SERVICE_
+ SERVICE_
+ SERVICE_ENDPOINT=http://
+ SERVICE_
+ S3_SERVICE_
+ KEYSTONE_
+ DEVSTACK_
+ ENABLED_
+ HEAT_API_CFN_PORT=
+ HEAT_API_PORT=
+ bash -x /home/stack/
+ ADMIN_PASSWORD=
+ SERVICE_
+ export SERVICE_
+ SERVICE_
+ export SERVICE_ENDPOINT=http://
+ SERVICE_ENDPOINT=http://
+ SERVICE_
++ get_id keystone tenant-create --name=admin
+++ awk '/ id / { print $4 }'
+++ keystone tenant-create --name=admin
Unable to communicate with identity service: (403, 'Forbidden'). (HTTP 400)
++ echo
+ ADMIN_TENANT=
++ get_id keystone tenant-create --name=service
+++ awk '/ id / { print $4 }'
+++ keystone tenant-create --name=service
Unable to communicate with identity service: (403, 'Forbidden'). (HTTP 400)
And you can see that it carries on after the error blindly continuing. keystone exited non-zero, so stack.sh can/should have noticed.
Changed in devstack: | |
status: | New → Triaged |
importance: | Undecided → Low |
Another example today; nova-api-os-volume was removed, but not from setup.py, and setup.py then fails to install all the subsequent scripts, but devstack carries blithely on...