toml: update the alexcrichton and BurntSushi repos to their successors toml-rs, and toml-test, record new exceptions

[spytheman]

Fix #20933 .

[spytheman]

@larpon, sorry for the noise. I realized too late, that the successor repo for alexcrichton, can be used directly, since it also contains the commit 499e8c4, which we need.

I think the PR should be stable enough for review now.

[larpon]

Hi @spytheman thanks for taking care of this. The changes look fine to me - the CI job definitely looks a lot nicer now. Did you have a specific reason to not rename burntsushi/alexcrihton, if those tests are now the official tests?

I only have one issue locally:

All separate tests (alexcrichton_toml_rs_test.v, burntsushi_toml_test.v iarna_toml_spec_test.v and large_toml_file_test.v) pass on my system (EndeavourOS Arch/rolling) when run separately, but running:

  VTEST_TOML_DO_YAML_CONVERSION=1 VTEST_TOML_DO_LARGE_FILES=1 ./v test vlib/toml

... locally gives me this:

Note the e vs. E I'm not sure if that is because I'm on newer versions of the software stack or what 🤔

I remember we had to do some shoehorning / tweaking of some test data/results because of the YAML conversions I'm not sure if this might be one of those 🤔

[larpon]

From https://github.com/vlang/v/pull/21152/files#diff-f9e436e6d2d57cf88c41f996f2c4be49302cc0559cd810f254cc7fcb7536bae1R343-R391
EDIT
From

v/vlib/toml/tests/iarna_toml_spec_test.v

Lines 343 to 381 in 59a5a79

	ast.Number {
	if value.text.contains('inf') {
	mut json_text := value.text.replace('inf', '1.7976931348623157e+308') // Inconsistency ???
	if skip_value_map {
	return '${json_text}'
	}
	return '{ "type": "float", "value": "${json_text}" }'
	}
	if value.text.contains('nan') {
	mut json_text := 'null'
	if skip_value_map {
	return '${json_text}'
	}
	return '{ "type": "float", "value": "${json_text}" }'
	}
	if !value.text.starts_with('0x')
	&& (value.text.contains('.') || value.text.to_lower().contains('e')) {
	mut val := '${value.f64()}'.replace('.e+', '.0e') // json notation
	if !val.contains('.') && val != '0' { // json notation
	val += '.0'
	}
	if skip_value_map {
	return '${val}'
	}
	return '{ "type": "float", "value": "${val}" }'
	}
	v := value.i64()
	// TODO: workaround https://github.com/vlang/v/issues/9507
	if v == i64(-9223372036854775807 - 1) {
	if skip_value_map {
	return '-9223372036854775808'
	}
	return '{ "type": "integer", "value": "-9223372036854775808" }'
	}
	if skip_value_map {
	return '${v}'
	}
	return '{ "type": "integer", "value": "${v}" }'
	}

So it might be a tool update that might have fixed/changed the scientific notation.

Also seems we've closed the i64(-9223372036854775807 - 1) special case: #9507
But it's probably better to look into that in a separate clean up

[spytheman]

I could not get the problem to show up locally @larpon , and it seems it is also passing on the CI. I suspect that the used jq version may be the culprit. My local copy is jq-1.6.

Edit: on the CI, it is also jq-1.6.

[spytheman]

Did you have a specific reason to not rename burntsushi/alexcrihton, if those tests are now the official tests?

Just minimization of the commit diff. I did not want to have both renames and code changes in the same PR.