sys.FileSystem.exists and leading "/"

[deepnight]

I have a "bar" directory somewhere on the disk:

sys.FileSystem.exists("../foo/bar"); // returns true

sys.FileSystem.exists("../foo/bar/"); // returns false

[deltaluca]

I personally would say that's correct behaviour, whether it's what is intended or not I don't know and it doesn't appear to be specifically documented regarding / tails.

The reason I say it's correct, is that to me (and this can often be the case with utilities) `bar/`` is an incomplete path, since you're specifying access to the 'bar' directory, but not providing a file/sub-directory name

[Simn]

I see where you're coming from with that, but it wouldn't hurt to support this anyway for convenience. The intent is clear enough even with an extra slash at the end.

Which targets does this fail on?

[hughsando]

I think the title should be 'trailing' slash.
This one has got me several times on neko. It means you need to write the
function "removeTrailingSlashfRequired" and would be great if it was fixed.

On Tue, Jan 28, 2014 at 3:51 AM, Simon Krajewski
notifications@github.comwrote:

I see where you're coming from with that, but it wouldn't hurt to support
this anyway for convenience. The intent is clear enough even with an extra
slash at the end.

Which targets does this fail on?

Reply to this email directly or view it on GitHubhttps://github.com//issues/2577#issuecomment-33413532
.

[Simn]

Could someone test this across all sys targets?

[ncannasse]

Maybe an unit test would be nice here?

[Simn]

File system unit tests are a bit problematic for CI servers IIRC.

[ncannasse]

@Simn maybe we can check for an existing local dir, like "../tests/" ?

[Simn]

@andyli: Is it safe to have file system tests like this on Travis? If so, which directory could we use, given that the cwds might be different.

[andyli]

Yes, we can do file system test on Travis. We have full system right that we could even try sudo rm -rf / (no pw is needed) ;)

Currently the RunTravis script cd into the tests/unit directory before compiling/running the tests, except for the optimization tests, which it cd into tests/optimization.
FYI, $TRAVIS_BUILD_DIR stores the absolute path to the directory where the repository being built has been cloned.

Since we still run the tests manually sometimes, I would suggest only messing with a specific directory inside tests/unit.

[Simn]

Aren't the cpp tests executed from within the cpp directory?

[andyli]

The cpp binary output is indeed inside test/unit/cpp directory, but it is executed from outside (test/unit).
See https://github.com/HaxeFoundation/haxe/blob/development/tests/RunTravis.hx#L87

Same for java/c#.

[Simn]

That's strange, I've just committed the test in cb87fba. It fails for me locally, but travis doesn't seem to mind. Could this be an OS thing?