Revision r48113 introduced Procs to delay time-consuming generation of error messages. This is a good idea, but it doesn't work for me.
How to reproduce: Make a small change in lib/unicode_normalize/tables.rb. Any change should do; I changed "Q" to "QQ" in one instance. Run make install(-nodoc). Run ruby test/test_unicode_normalize.rb.
The result I get is as follows:
===============================================================================
Failure:
test_normalize_to_NFKD_from_source_with_nfkd(TestNormalize)
test/test_unicode_normalize.rb:38:in `block (2 levels) in generate_test_normalize'
test/test_unicode_normalize.rb:33:in `each'
test/test_unicode_normalize.rb:33:in `block in generate_test_normalize'
#<Proc:0x000006033229f0@test/test_unicode_normalize.rb:32>
<"Q"> expected but was
<"QQ">
diff:
? QQ
===============================================================================
The failure is expected (artificially generated). The problem is the line #<Proc:0x000006033229f0@test/test_unicode_normalize.rb:32>
The Proc that is supposed to be evaluated to generate a more detailed error message (showing codepoints of the characters involved) is just converted to a String.
For reference, I'm using the standard bundled gems minitest-5.4.1.gem, power_assert-0.1.4.gem, and test-unit-3.0.1.gem.
[29/47] TestNormalize#test_normalize_to_NFKC_from_NFC_with_nfkc = 0.01 s
1) Failure:
TestNormalize#test_normalize_to_NFKC_from_NFC_with_nfkc [/Users/nobu/src/ruby/trunk/src/test/test_unicode_normalize.rb:38]:
["0051"] expected but was ["0051", "0051"] on line 1232 (nfkc).
<"Q"> expected but was
<"QQ">.
This is one possibility. But I'm using the standard installation with the bundled gems (as written above: minitest-5.4.1.gem, power_assert-0.1.4.gem, and test-unit-3.0.1.gem; see also #10380). Is there a problem with that?
Third, test-all is very good to have, but it's too slow when I know which tests I need to run.
I see.
You run the test just as ruby test/test_unicode_normalize.rb?
You need to run it via test/runner.rb, i.e., ruby test/runner.rb test_unicode_normalize.rb.
Third, test-all is very good to have, but it's too slow when I know which tests I need to run.
I see.
You run the test just as ruby test/test_unicode_normalize.rb?
Yes.
You need to run it via test/runner.rb, i.e., ruby test/runner.rb test_unicode_normalize.rb.
Thanks for the advice, it worked.
But why do we need a special runner.rb script just to run individual tests? Now that Unicode normalization functionality is part of Ruby itself, I want to create a gem for people who want to use it on older versions, but the need for such a test runner script or the change from tool/unicode_norm_gen.rb to template/unicode_norm_gen.tmpl create unnecessary divergence.
I've released new test-unit gem. It support Proc as user_message.
Should I update version in gems/bundled_gems?
It seem that Hiroshi has done this at r48180, and I have updated my checkout, but I still get the same error. I'm not sure why, but could it be because now both gems get installed? This is what gets shown at the end of 'make install-nodoc'
Updated by 
Updated by 
Updated by