Fix tests to be string hash order independent

[c-blake]

For new Farm Hash-based hash(x:string).

[narimiran]

I haven't explored the reported error in detail, but it looks like it might be a missing import std/algorithm problem, to have sorted on seqs.

[c-blake]

std/algorithm is there right at the top of the file. I think it is more missing a < as the error message says. I pushed a commit to define one. Can you re-run the tests? Thanks.

[c-blake]

Another compile error, another git push to try to fix it. :-)

Needless to say, I am not running these tests locally (I am not set up to). Much appreciated if you could approve the workflow again.

[c-blake]

Ok. Naming fields with tuple[] is non-optional. Dumb commit message.

This time I at least confirmed compilation & functioning locally of

import std/algorithm
proc `<`(a, b: (string,string)): bool = a[0] < b[0]
echo @[("Header2", "value2"), ("Header2", "VALUE3"), ("Header1", "value1")].sorted

So, I have high hopes this version will work. Thanks for approving the workflow yet again!

[c-blake]

Ack. I just kept mis-reading compiler errors. This works for me locally and should work now if you approve the workflow yet again (famous last words, as they say!):

import std/[algorithm, assertions]
proc `<`(a, b: (string, seq[string])): bool = a[0] < b[0]
let data1 = @[("Header2", "value2"), ("Header2", "VALUE3"), ("Header1", "value1")]
let data2 = @[("Header2", @["value2", "VALUE3"]), ("Header1", @["value1"])]
assert data1.sorted == sorted(@[("Header2", "value2"), ("Header2", "VALUE3"),
                                ("Header1", "value1")])
assert data2.sorted == sorted(@[("Header2", @["value2", "VALUE3"]),
                                ("Header1", @["value1"])])

[narimiran]

Heh, we're back at Error: type mismatch: got <seq[string], seq[string]>

[c-blake]

I don't know what's wrong now. That above snippet really does run fine locally on nim-devel. I could add a <(a,b: seq[string]), but I'm not sure if that's ok. Do you know what's wrong? Is this some known gotcha with the test suite thing? Should I lift that < to outermost scope?

[c-blake]

Well, with yesterday's final commit (lifting < to outermost scope as suggested in the last comment here), the checks pass now on nim-devel (& older).

There may be other considerations in tagging a new Chronos release so as to pass the main Nim CI, but as far as this test failure is concerned, I think the work is done and this PR is ok to merge. Let me know if this is not the case.

Thanks