Regression tests for the REPL

[bstrie]

rusti is great for people who are learning the language, but it's quite fragile and the devs themselves don't have much cause to use it. Regression tests are needed to keep it intact. I have no idea how to go about doing this.

[jdm]

Running cat input | rusti causes rusti to execute the code in input line by line. The tests could be pairs of input and expected output files and passing could be defined as diff output expected returning no output.

[bstrie]

True, though we can't assume that rusti is installed so it would have to be something like cat input | $RUST_DIR/stage2/bin/rusti

[jbclements]

On Mar 21, 2013, at 9:23 AM, Ben Striegel wrote:

True, though we can't assume that rusti is installed so it would have to be something like cat input | $RUST_DIR/stage2/bin/rusti

Maybe I missed something. Isn't it possible to construct test cases for rusti that don't go through stdin/stdout? It seems to me that the io can be tested elsewhere.

John Clements

[bstrie]

I don't think a new tag for rusti is needed right now. A-tools is already quite sparse, and afaik the planned feature set for the REPL is small.

[lkuper]

I just came across this discussion while in the process of tagging issues with A-rusti. There are 9 issues tagged with it at the moment, which seems like enough to justify having its own tag, although admittedly this is an after-the-fact sort of justification.

But, in any case, I think rusti deserves to get a lot more love, and part of that is having a tag; it'll help people remember that rusti is a thing, or, at least, it won't hurt.

+1 for regression tests for rusti, as well.

[lkuper]

@astrieanna and I can claim this bug.