Fix utf8 tests

[guillep2k]

modules/charset/ tests have been failing randomly. My guess, once again, is the LOCALE on the CI server might be producing diferent compilation results when the compiler parses the strings.

This time around I've eliminated all but the simplest string literals, e.g.:

res = ToUTF8("áéíóú")

changed into

res = ToUTF8(string([]byte{0xef, 0xbb, 0xbf, 0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3,
		0xc3, 0xba}))

I hope this is the solution. Otherwise, I'd recommend just removing all the UTF8 test.

[codecov-io]

Codecov Report

❗ No coverage uploaded for pull request base (master@73f7e82). Click here to learn what that means.
The diff coverage is n/a.

@@            Coverage Diff            @@
##             master    #8192   +/-   ##
=========================================
  Coverage          ?   41.58%           
=========================================
  Files             ?      488           
  Lines             ?    65211           
  Branches          ?        0           
=========================================
  Hits              ?    27116           
  Misses            ?    34607           
  Partials          ?     3488

Impacted Files	Coverage Δ
modules/charset/charset.go	69.56% <ø> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 73f7e82...6da827b. Read the comment docs.

[lunny]

how about to set a locale on tests?

[guillep2k]

how about to set a locale on tests?

@lunny As I understand it the problem is the compiler's locale, not the actual code. When the compiler parses the strings to generate code, it's got rules and defaults to deal with invalid data that are (or should be) based in the locale. What we could do is to set a LOCALE in the Makefile to force one during compilation.

Anyway, with the changes I've made in this PR that shouldn't be necessary, as all the strings are generated from bytes in the code and I've attempted to isolate all the conflicting strings as much as possible. The compiler's locale should no longer affect the generated code nor the test run.

[techknowlogick]

ping LG-TM bot