Add mesh reader test for trusses

[jtveiten]

The mesh part is now read. Need to do the model part later, keyword density isn't recognized yet

[coveralls]

Coverage remained the same at 100.0% when pulling cd3e8e0 on jtveiten:master into a9677c7 on JuliaFEM:master.

[ovainola]

Issues discussed in chat must be addressed

[ovainola]

et22sfse.inp file is copyrighted by Dassault Systems, file needs to be fetch from external location

[ahojukka5]

We are not sure if it is ok to add that test file into the repository, it probably is copyrighted by Dassault. I implemented a new function download_abaqus which can be used to download files on the fly before testing, requiring that one has set ABAQUS_DOWNLOAD_URL to point someplace where the files can be downloaded. See PR #29, and here.

Could you do the following changes:

• rebase from newest master to get download_abaqus
• remove file et22sfse.inp from repository
• use fn = abaqus_download("et22sfse.inp") to download the file during the testing. fn will be the path for the downloaded file.

If you want to do local testing, you must do something like ENV["ABAQUS_DOWNLOAD_URL"] = "https://my_models.com/v6.14/books/eif" before using the function, otherwise, it returns nothing.

Also, consider squashing changes to keep version history clean. This also prevents the file et22sfse.inp from being written into the repository at any point. Here is a quite good explanation how squashing works.

[ahojukka5]

Looks that my test is breaking further tests. I have to fix it.

[jtveiten]

First of all, apologies for not looking at the terms and conditions.
I have done the changes, but I wasn't sure I got the squashing right after some trys, so I haven't done that yet.
The test runs locally on my machine as far as I can see?

[ahojukka5]

It was another test messing up environment variables and not returning the original state. If you now rebase again for newest master this should work.

[jtveiten]

If I run it locally after setting the env. variable, including the reader and including runtests.jl, it seems to succeed. But if do an include("runtests.jl") again, it fails on my machine

[ahojukka5]

One more time, rebase for the newest master.

[ahojukka5]

Hmm https://docs.travis-ci.com/user/pull-requests/#Pull-Requests-and-Security-Restrictions

[jtveiten]

Do I need to protect the test by checking for an empty string returned from abaqus_download?
This will work once it the code is in master? Or a different approach on reistering filenames?

[ahojukka5]

Well the test probably should work, if branch is under JuliaFEM/AbaqusReader.jl, let see what happens when I opened a new PR based on this change. PR #32.

[jtveiten]

So that worked. But it is unfortunate that the forks don't build.

[ahojukka5]

Ok this is now merged in PR #32.

[ahojukka5]

Yes, but there is a good reason why environment variables are not set for pull requests coming outside of the organization. Someone could do something like println(ENV["PASSWORD"]) and then read the output from Travis CI log...