fix 3 minor bugs in joinPath (see #13455)

[a-mr]

this fixes 3 wrong cases for joinPath:

"" / "" == "/", should be ""

"/" / "" == "//", should be "/"

"/" / "/a/b/c" == "//a/b/c", should be "/a/b/c"

I don't fix behavior like "a" / "" == "a/" since it's specified so in documentation hence changing it might break compatibility, though I admit @timotheecour is right in #13455 that it's strange.

[timotheecour]

thanks! a definite improvement

I don't fix behavior like "a" / "" == "a/" since it's specified so in documentation hence changing it might break compatibility, though I admit @timotheecour is right in #13455 that it's strange.

this can be debated/fixed in subsequent PR's, however your PR introduces a new inconsistency; which I'm sure can be addressed in your PR: when head ends with / it strips /; when it doesn't end with / it adds it

echo joinPath("././foo/.//","././") # pre-existing inconsistency
echo joinPath("foo/","") # new inconsistency that wasn't there before this PR
echo joinPath("foo","")

after PR:

foo # pre-existing inconsistency
foo # new inconsistency
foo/

before PR:

foo
foo/
foo/

so how about changing the logic so that joinPath("foo/","") returns foo/, to at least avoid adding an inconsistency; and in future PR we can revisit this and consider making all of these return foo (that debate should happen somewhere else)

[a-mr]

wow, well-spotted, thanks!
Inconsistency is fixed.

[Araq]

Needs a changelog entry and .since support.

[a-mr]

I'm not sure that I understood the point of pragma .since.

[timotheecour]

@a-mr your tests are failing because you need include system/inclrtl for since, HOWEVER:

I'm not sure that I understood the point of pragma .since.

I agree with you, IMO since shouldn't be used in this particular case; the old behavior was a just bug, not something useful/justifiable; code that depends on compiler/stdlib bugs to work is bound to break. --useVersion:1.0 should be irrelevant here.

we should backport this otherwise user code will need to be sprinkled with when (NimMajor, NimMinor) >= (1,1) to support both nim 1.0.X and nim 1.1.

So I suggest to not use since in this case.

[a-mr]

Thank you for the great explanation. I removed since.