Explore deleting rl.exe

[rhuanjl]

The folder bin/rl contains source and build files for rl.exe this is a windows only test runner application, it is currently built and used by the CI on windows.

BUT I can't see anything it does that test/runtests.py can't do.

Unless anyone has an objection?
We should:

1. update test/runtests.py to be fully functional on windows (if it isn't)
2. update test/jenkins.testone.cmd to call runtests.py instead of runtests.cmd
3. Delete runtests.cmd (wrapper script around rl.exe)
4. Delete bin/rl and all references to it

The one key reason I could see not to do this is if rl.exe is significantly faster than runtests.py (if so we should contemplate getting rid of runtests.py instead though that would be more work) - best way I can see to test this is to update runtests.py to handle windows then change the CI to run on it and see what happens.

[rhuanjl]

#6592 gives me confidence that runtests.py can match rl.exe performance.

Though I've also noticed that runtests.cmd currently has a lot of alternate test flags that we're not using at the moment BUT perhaps should be using at some point - so I think I should move all of those into runtests.py before runtests.cmd is deleted.

[rhuanjl]

Removed Help Wanted label as I've started working on this.

[rhuanjl]

#6596 has disabled rl.exe in favour of runtests.py

However I want to port the various extra testing options from runtests.cmd to runtests.py before following up with deleting rl.exe and runtests.cmd

[rhuanjl]

Also need to update readme.md and contributing.md to point to using runtests.py for testing on windows.

[ppenzin]

Right now there seem to be slight result differences between runtests.cmd and runtests.py on Windows, I think that is due to how rl.exe handles whitespace. The cmd file has a more intuitive interface when used with VS build - it does not need a path to executable. I am on the fence about moving that to runtests.py - CMake builds does not really need it as CMake can set the path. I think maybe we need to make runtests.cmd invoke runtests.py for the time being to preserve some of the VS-only behavior.

I can look into it, since I use VS.

Also, there is a reference to the cmd file which is not valid anymore in this file: https://github.com/chakra-core/ChakraCore/blob/master/test/ci.testone.cmd#L25

[rhuanjl]

When you say result difference do you mean how it's displayed in the command line OR actual different number of passes/tests run?

[ppenzin]

It actually fails on constructor test, and when I diff the output with baseline the difference is windows newline characters. However I am not sure this should be a reason to keep fixing the tests, maybe just change .cmd file to call the Python script.

[rhuanjl]

runtests.py has logic to normalise the baseline line endings whereas rl.exe does not; for this reason they all used to be windows style (CRLF) - clearly I forgot to run unix2dos on that baseline when checking in #6620

If we're not deleting rl.exe yet should probably fix it.