rosie: restore rosie ws

[oliver-sanders]

NOTE: As sqlalchemy was installed to system Python rather than Travis Python some tests wern't being run (due to missing dependencies). These have been fixed.

• Restore files removed in 4004944.
• Re-jig .travis/now with a new RI_PATH for setting PATH in rose_site_init in env -i environments.
• Fix tests.

[wxtim]

What is there looks ok. Is it sufficient?

[matthewrmshin]

Travis CI build failing due to lack of request module?

[oliver-sanders]

By adding the Rosie dependencies we are now activating tests which might not have been run since Python2.

Rosie tests are falling over in a peculiar way. It would appear svn_post_commit is being run under Python3.5 (OS provided) rather than 3.7 (Travis installed) where it then fails due to the missing sqlalchemy dependency which is installed in the 3.7 stack.

[oliver-sanders]

In #2288 sqlalchemy was installed to system Python via apt-get.

• Added rose_init_site updating PATH to include the dirs including python and rose in the event that $HOME is not defined (commit_hook scripts).

not working yet :(

[oliver-sanders]

The tests should pass this time:

• For the moment I've moved pygraphviz imports out the way awaiting complete removal.
• I've skipped a failing test which is punted to rosie disco: restore "slash at the end" test #2357

[sadielbartholomew]

I've skipped a failing test which is punted to #2357

I had a look-see at this, & it is failing for quite a significant reason: the rosie-disco web files are not there!

Because all other t/rosie-disco/ sub-tests were testing on ?format-json endpoints, the only two tests that were checking for rendered pages, which use the HTML templates & CSS & JS files from the lib/html/ directory in 2019.XX & previous versions, were the sub-tests for the root index page! And the second of those:

https://github.com/oliver-sanders/rose/blob/a2309eb25e6768d9c0628d9da75e1f3adb9f91b5/t/rosie-disco/00-basic.t#L65-L70

was just testing for a re-direct to the equivalent with the trailing slash, so it passed. But it redirected from a 500 Internal Sever Error, as was being returned for the redirected page causing that one sub-test failure.

Presumably at some point the web files required were removed from master. Though we no longer need the subset of them designed for Rose Bush, we still want them for the updated Python 3 Rosie Disco. I will put in a PR to this branch which re-adds the necessary directories (leaving out obvious Rose Bush files), & removes the comment characters on the currently-failing sub-test which should then pass (Travis, based on the local test battery pass I then see).

[oliver-sanders]

Thanks Sadie, hopefully this should now pass.

[matthewrmshin]

Travis CI unit test failing - need to filter out lib/html/ hierarchy?

[sadielbartholomew]

Ah, & naturally those web files come a bundle of issues as flagged also by Codacy. Can we get them ignored & (re-)raised as an Issue? I can't recall quite how we dealt with those previously...

[matthewrmshin]

On Codacy. We should filter out the external libraries from the analysis. Some of our own stuffs can be fixed. The others should be flagged for ignore.

[matthewrmshin]

Branch in conflict.

[sadielbartholomew]

@oliver-sanders what is the state of this PR, so I know when to review? Is the plan to address those Codacy points? Also I have kicked Travis CI but across a number of tries it keeps returning an error from a syntax error in .travis/now.

[oliver-sanders]

PR was still ready for review, tests broken by most recent conflict resolution.

[matthewrmshin]

Unit test broken by style check?

[oliver-sanders]

The three codacy issues can be ignored:

• The Jinja2 autoescape comment is just wrong as this is set on the following line
• The input comments are not relevant to Python3, Tim has ignored these issues in Another rose package branch #2360

[matthewrmshin]

@wxtim Please perform a final sanity check. The Codacy/Bandit issue can be unflagged in your branch?