Fix some issues with newlines in fixes on Windows

[ZedThree]

When I deduplicated building the binaries, it also stopped us building on different platforms.

And it looks like we were never actually testing on those platforms!

Windows failures seem pretty simple: paths are using \ instead of / as expected in the snapshots. Can probably do something simple to fix those in the tests.

Mac failures more puzzling, looks like per-file-ignores not working quite as expected?

[LiamPattinson]

For the Mac problem, perhaps it's this line:

fortitude/fortitude/tests/check.rs

Line 699 in ea2d6b8

.arg(path)

We're using .current_dir(path) and also .arg(path), which might be confusing things. I think removing .arg(path) might fix it, but without a Mac in my hands I can't really check.

[ZedThree]

We're using .current_dir(path) and also .arg(path), which might be confusing things. I think removing .arg(path) might fix it, but without a Mac in my hands I can't really check.

Nice, good catch!

Failures on Windows now are due to line-ending handling -- I think these are actual bugs, not just issues in the tests

[ZedThree]

Hmm remaining failure is this:

---- rules::style::tests::rules::rule_incorrectspacebeforecomment_path_new_s102_f90_expects stdout ----
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Snapshot Summary ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Snapshot file: fortitude\src\rules\style\snapshots\fortitude__rules__style__tests__incorrect-space-before-comment_S102.f90.snap
Snapshot: incorrect-space-before-comment_S102.f90
Source: D:\a\fortitude\fortitude:42
───────────────────────────────────────────────────────────────────────────────
Expression: diagnostics
───────────────────────────────────────────────────────────────────────────────
-old snapshot
+new results
────────────┬──────────────────────────────────────────────────────────────────
   11    11 │ ℹ Safe fix␊
   12    12 │ 1 1 | ! This is fine␊
   13    13 │ 2 2 |  ! This isn't but we'll let it pass␊
   14    14 │ 3   |-module mymod! This should be given two extra spaces␊
   15       │-  3 |+module mymod  ! This should be given two extra spaces␊
   16       │-4 4 |   implicit none ! This should be given one extra space␊
   17       │-5 5 |   private  ! This is fine␊
   18       │-6 6 |   ! This is fine␊
         15 │+  3 |+module mymod  ! This should be given two extra spaces␍
         16 │+  4 |+␊
         17 │+4 5 |   implicit none ! This should be given one extra space␊
         18 │+5 6 |   private  ! This is fine␊
         19 │+6 7 |   ! This is fine␊
   19    20 │ ␊
   20    21 │ ./resources/test/fixtures/style/S102.f90:4:17: S102 [*] need at least 2 spaces before inline comment␊
   21    22 │   |␊
   22    23 │ 2 |  ! This isn't but we'll let it pass␊
┈┈┈┈┈┈┈┈┈┈┈┈┼┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈
   32    33 │ 1 1 | ! This is fine␊
   33    34 │ 2 2 |  ! This isn't but we'll let it pass␊
   34    35 │ 3 3 | module mymod! This should be given two extra spaces␊
   35    36 │ 4   |-  implicit none ! This should be given one extra space␊
   36       │-  4 |+  implicit none  ! This should be given one extra space␊
   37       │-5 5 |   private  ! This is fine␊
   38       │-6 6 |   ! This is fine␊
   39       │-7 7 | ! As is this
         37 │+  4 |+  implicit none  ! This should be given one extra space␍
         38 │+  5 |+␊
         39 │+5 6 |   private  ! This is fine␊
         40 │+6 7 |   ! This is fine␊
         41 │+7 8 | ! As is this
────────────┴──────────────────────────────────────────────────────────────────

and I cannot see anything obvious. I would've guessed that this could happen if we were writing a newline, but I'm reasonably sure we're explicitly not.

It might be quite tricky to debug this without a Windows env!

[LiamPattinson]

That's really strange. Am I reading it correctly in that it it's adding an extra newline in Windows?

[ZedThree]

Sort of -- it's changing the existing line ending from \n (unix) to \r\n (DOS line ending). The characters in the output are really small, but you can see CR/LF.

My understanding is that we're replacing the comment node upto but not including the line ending, so we shouldn't be writing one back out. I don't think we explicitly write new lines anywhere?

[LiamPattinson]

I think this might be similar to an issue we saw with end module etc, where the node included a newline character. Oddly, I think tree-sitter was adding a trailing \r via utf8_text. I wasn't able to completely figure out what was going wrong and why, but I was able to fix it up by simplifying the edit. I also fixed up a couple other Windows issues, and I'll raise a PR once it's all finalised.

[ZedThree]

Nice, thanks for the extra fixes! The insertion is much simpler.

I've investigated further, and the tree-sitter grammar consumes the \r character into comments -- that might be something we have to watch out for in future. Or work out how to fix it in the grammar.