Issues with the official test submodule

[RebelusQuo]

I want to try my hand at Javascript and C# ports (in that order), currently working on rigging a test suit based on the test submodule.

However, firstly I noticed a number of json and nt files which simple have file names as their text content (e.g. .../dump_in.json having the content load_out.json, which is obviously not valid Json), and secondly there's stale issues and PRs in that repo which are small enough to be easy to address even if this library is on low priority.

I feel this needs to be handled for me to continue without too much friction. I could just copy the test suit, but that wouldn't feel proper in light of what seems to be the intention of a separate shared repo.

(I posted the issue in this repo as it seems to be more active)

[RebelusQuo]

so... I just noticed these files have a different icon in the github web interface. They're *nix shortcuts? cries in Windows

[KenKundert]

Sorry, I am having a little trouble deciphering your messages. As I understand it, you are on a Windows machine and the tests are broken because Windows does not interpret the symbolic links properly. Is that correct? There is some discussion about refactoring the testcases, at which time we could eliminate the symbolic links, however that is not currently scheduled. In the mean time, you can write a script that runs over the test cases checking the dump files, and if they are links, replace them with the corresponding load file.

[RebelusQuo]

Sorry for my swedish-english, it does get in the way sometimes :) Ok, so they are indeed symbolic links. I think I see the intention now, they set up a reversed load/dump case?

On my platform I just get an ordinary file which hold a file name but I could set up something in code to handled that. Although I should check if my git4win install has a setting I should fiddle with. I'll keep you posted.

[RebelusQuo]

I decided to go the script route. While git4win has an install option for symbolic link support, this is in turn based on the native windows implementation. However, there's some limits and avoiding the need for UAC elevation requires a number of steps. I'd much rather link a (relatively) simple script in my readme than the trouble of navigating the insides of Windows itself.

Btw, in case someone is interested, the scripts I found are from https://stackoverflow.com/a/16754068

So, is there anything you can say about this refactoring? Major changes, or relatively minor? Getting an outline could be helpful as preparation.

[KenKundert]

An initial version of the refactored tests is available here.