test_writer_indent_cdata fails on linux with stray "13" items

[decathorpe]

Running the test suite for quick-xml 0.23.0 on Fedora Linux with stray "13" items.
It looks like this test discriminates between windows and not(windows) for some reason:
https://github.com/tafia/quick-xml/blob/master/tests/unit_tests.rs#L264

And it seems that the reason to discriminate between windows and not(windows) might no longer be there?

Environment: Fedora Linux 36 on x86_64 with Rust 1.62.0

thread 'test_writer_indent_cdata' panicked at 'assertion failed: `(left == right)`

Diff < left / right > :
 [
     60,
     63,
     120,
     109,
     108,
     32,
     118,
     101,
     114,
     115,
     105,
     111,
     110,
     61,
     34,
     49,
     46,
     48,
     34,
     63,
     62,
>    13,
     10,
     60,
     106,
     111,
     98,
     115,
     62,
>    13,
     10,
     32,
     32,
     32,
     32,
     60,
     106,
     111,
     98,
     62,
>    13,
     10,
     32,
     32,
     32,
     32,
     32,
     32,
     32,
     32,
     60,
     106,
     111,
     98,
     105,
     100,
     62,
     60,
     33,
     91,
     67,
     68,
     65,
     84,
     65,
     91,
     48,
     48,
     100,
     52,
     54,
     101,
     52,
     52,
     57,
     52,
     101,
     49,
     93,
     93,
     62,
     60,
     47,
     106,
     111,
     98,
     105,
     100,
     62,
>    13,
     10,
     32,
     32,
     32,
     32,
     60,
     47,
     106,
     111,
     98,
     62,
>    13,
     10,
     60,
     47,
     106,
     111,
     98,
     115,
     62,
 ]

[dralley]

There's a lot of reasons I'm going to be very happy about the API redesign, and one of them is not having to look at lists of digits.

And it seems that the reason to discriminate between windows and not(windows) might no longer be there?

It's definitely possible. Will try to take a look later

[Mingun]

It seems that it is depend on the way of how git's checkout test XML files (EOL mode)

[decathorpe]

That sounds like it's safe to ignore this specific test failure?

[Mingun]

I would like to change the git rules to treat all files in test/documents directory as binary and remove from tests any platform-specific code

[Mingun]

Strange, but repository already contains .gitattributes file, but on my windows machine line endings of XML files still converted to CRLF...

quick-xml/.gitattributes

Lines 1 to 2 in c145576

	# Unit tests assume that all xml files have unix style line endings
	*.xml text eol=cr

[Mingun]

I found that

• repository contains /tests/documents/* files with LF endings. Checked that by checkout on linux where core.autocrlf=false
• checkout on linux uses LF ending in working copy
• checkout on Windows uses CRLF ending in working copy, because of my local git config core.autocrlf=true, which is normally fine for files with the code
• *.xml text eol=cr does nothing, according to this answer. I suspect that *.xml text eol=lf should be used
• *.xml text eol=lf does not change situation on Windows immediately, you should run git add --renormalize . to tell git that LF -> CRLF change due to git config core.autocrlf=true in your working copy is an actual change now and then git reset --hard to remove that changes
• on Windows utf16le.xml, utf16be.xml and sample_5_utf16bom.xml may be broken on checkout because of incorrect LF -> CRLF conversion, which inserts bytes assuming UTF-8 input. As a result, file content was changed:
  

@decathorpe, actually, it is strange, that test is failed on Linux. I would assume that you have git config core.autocrlf=true and your working copy have CRLF endings, which is unusual for Linux. Because our CI works on Ubuntu, I assume that your problem due to that reason and #454 should fix it. Could you check?

[decathorpe]

I did not run the tests from a git checkout, but from the canonical sources that are uploaded to crates.io. So the contents of those files might depend on which system "cargo publish" was run on.

[Mingun]

Yes, 0.23.0 I've built on Windows. So it seems a good theory