Convert CI from M3 to M4

See merge request !1

https://github.com/autolab/Tango/issues/153

Here, using the gitlab-ci-user which the docs say is (mostly)
appropriately for public projects (which Tango is currently).
See notes here for discussion and other options.
https://docs.gitlab.com/ee/ci/docker/using_docker_build.html

vmms/Dockerfile_wsuv_* are also slightly improved with
 - fewer layers
 - pulling from wsuv repos instead of cmu's

Bugfix: restful-tango/server.py truncates files

See merge request autolab-project/m3/Tango!6

localDocker.py now works on ubuntu 18.04 image (see note)

See merge request autolab-project/m3/Tango!3

wallaces authored 6 years ago and matthew.bourgoine committed 6 years ago

the original invocation of the autodriver redirected output
as cmd &> outfile; but this fails in a autograder built on
ubuntu 18.04 with a message:

An error occurred while parsing the autoresult returned by the Autograder.

Error message: 795: unexpected token at '---'

The fix applied here (do: cmd > output 2>&1 ) is also suggested by
issue 153 on cmu's tango repo:
https://github.com/autolab/Tango/issues/153

Add -p flag to mkdir to Dockerfile to ensure idempotency

Fix regression during file upload

Regression caused by commit 05e4d0f3. Files are
copied from a temp file to the final destination as the file has not been fully
uploaded and flushed. That yielded to grading corrupted or empty files.

* use tornado streaming for uploads

use tornado.web.stream_request_body for the upload handler to avoid
reading the entire upload into a memory buffer *in the IOLoop*. Instead,
hash the content as it's streamed in and store it in a temporary file under
COURSELABS. Then the upload business logic can just rename the temporary file
over the destination if the upload is accepted. This significantly improves
latency when large files are uploaded

* Remove tempfile if it's a duplicate

Remove the temporary upload file if it's a duplicate instead of leaving it
in the temp directory.

Ensures server arg in CLI is not necessary. Makes ramp up process easier

Changes:
-  Allow Autolab front-end to specify the AWS credentials that are used in auto-grading
-  Back-compatibility preserved (new features are effective only when accessKeyId and acceessKey is passed in)
-  change restful clients accordingly

* add EC2_USER_NAME to config template file

* ec2 VMM bug fixes and resolve hard-coded user name issue

* fixed typo and indent error

running Docker container