Bug #20929: TestTime have an assertion different from current implementation.

• Status: Closed
• Assignee: -
• Target version: -
• Backport: 3.1: UNKNOWN, 3.2: UNKNOWN, 3.3: UNKNOWN

[YO4 (Yoshinao Muramatsu)]

test/ruby/test_time.rb have following assersion function.

  def assert_zone_encoding(time)
    zone = time.zone
    assert_predicate(zone, :valid_encoding?)
    if zone.ascii_only?
      assert_equal(Encoding::US_ASCII, zone.encoding)
    else
      enc = Encoding.default_internal || Encoding.find('locale')
      assert_equal(enc, zone.encoding)
    end
  end

In current implementation, Time#zone are returned in US_ASCII or locale encoding, which does not seem to take into account the default_internal.

C:\>ruby -e "puts Time.now.zone"
東京 (標準時)

C:\>ruby -e "puts Time.now.zone.encoding"
Windows-31J

C:\>ruby -EWindows-31J:UTF-8 -e "puts Time.now.zone"
東京 (標準時)

C:\>ruby -EWindows-31J:UTF-8 -e "puts Time.now.zone.encoding"
Windows-31J

[nobu (Nobuyoshi Nakada)]

• Description updated (diff)

Indeed, that assertion is incorrect.

But the locale is not the correct/expected encoding always on Windows.

For instance, in Japanese edition, tm_zone is always CP932.

> chcp.com 932
現在のコード ページ: 932

> ruby -e "p Encoding.find('locale'), (z = Time.now.zone), z.encoding"
#<Encoding:Windows-31J>
"\x{938C}\x{8B9E} (\x{9557}\x{8F80}\x{8E9E})"
#<Encoding:Windows-31J>

Even when locale (active codepage) is changed.

> chcp.com 437
Active code page: 437

> ruby -e "p Encoding.find('locale'), (z = Time.now.zone), z.encoding"
#<Encoding:IBM437>
"\x93\x8C\x8B\x9E (\x95W\x8F\x80\x8E\x9E)"
#<Encoding:IBM437>

And of course regardless the internal encoding.

> ruby -Ecp932 -e "p Encoding.find('locale'), (z = Time.now.zone), z.encoding"
#<Encoding:IBM437>
"\x93\x8C\x8B\x9E (\x95W\x8F\x80\x8E\x9E)"
#<Encoding:IBM437>

I don't think there is the API to obtain what codepage it is encoded in.
Maybe we should use the W API and encode it in UTF-8 ranter than the locale.
@usa (Usaku NAKAMURA), what do you think?

[usa (Usaku NAKAMURA)]

Maybe we should use the W API and encode it in UTF-8 ranter than the locale.

agreed.

[nobu (Nobuyoshi Nakada)]

• Status changed from Open to Closed

Applied in changeset git|78762b52185aa80ee55c0d49b495aceed863dce2.

---

[Bug #20929] Fix assert_zone_encoding

The default internal encoding is not taken into account to encode
timezone name.

[YO4 (Yoshinao Muramatsu)]

Thank you for your response.

Regarding Time#zone encoding, I am experimenting with it in my branch https://github.com/YO4/ruby/tree/tzname_utf8.
I found this issue in my research for that.

At present, the change to utf-8 causes the following error.

>ruby -e 'puts "タイムゾーン:#{Time.now.zone}"'
-e:1:in '<main>': incompatible character encodings: Windows-31J and UTF-8 (Encoding::CompatibilityError)

To resolve this, other strings must also be in UTF-8 encoding.
I think it would be preferred that strings with Unicode code ranges also have UTF-8 encoding. OS-derived strings, excluding I/O content, seem to meet that requirement.

Of course, this matter should be discussed in another issue.
Thanks.

[YO4 (Yoshinao Muramatsu)]

We have changed Time#zone encoding test code to not consider internal_encoding at 78762b5,
but a document of Encoding.default_internal explicitly stated that Time#zone was affected by it.
Do we need to reconsider?

Github PR#12409 is the change for Time#zone to respect Encoding.default_internal.

78762b5
Document(3.3)
Document(master)
Github PR#12409