Some intl tests fail due to narrow non-breaking spaces

[nielsdos]

Description

The following tests:

IntlDateFormatter, calendars and time zone [ext/intl/tests/dateformat_calendars_variant3.phpt]
IntlDateFormatter::create() with default date and time types [ext/intl/tests/dateformat_create_default.phpt]
IntlDateFormatter::formatObject(): IntlCalendar tests [ext/intl/tests/dateformat_formatObject_calendar_variant5.phpt]
IntlDateFormatter::formatObject(): DateTime tests [ext/intl/tests/dateformat_formatObject_datetime_variant5.phpt]
datefmt_format_code() and datefmt_parse_code() [ext/intl/tests/dateformat_format_parse_version2.phpt]
datefmt_set_timezone_id_code() icu >= 4.8 [ext/intl/tests/dateformat_set_timezone_id3.phpt]
IntlDatePatternGenerator::getBestPattern() [ext/intl/tests/datepatterngenerator_get_best_pattern.phpt]
MessageFormatter::format(): DateTime accepted to format dates and times [ext/intl/tests/msgfmt_format_datetime.phpt]
MessageFormatter::format(): simple types handling with numeric strings [ext/intl/tests/msgfmt_format_simple_types_numeric_strings.phpt]
ResourceBundle constructor bundle accepts NULL for first two arguments [ext/intl/tests/resourcebundle_null_mandatory_args_variant2.phpt]

Break on my system (INTL_ICU_VERSION == 72.1)

For example, ext/intl/tests/dateformat_calendars_variant3.phpt resulted in the following diff:

========DIFF========
001+ string(16) "h:mm:ss a zzzz"
001- string(14) "h:mm:ss a zzzz"
     string(13) "HH:mm:ss zzzz"
========DONE========

The reason the string is 16 long instead of 14 is because on my system the space between "a" and "zzzz" is a UTF-8 "NARROW NO-BREAK SPACE". A "NARROW NO-BREAK SPACE" is 3 bytes long which explains the difference.
This "NARROW NO-BREAK SPACE" is the root cause of all the failing tests I posted, and I've also seen this one happen on CI too.

PHP Version

8.1.14 and higher

Operating System

Linux 6.1.3

[Girgias]

Is this an ICU change? In which case we need to duplicate the test and add some version constraint in the SKIPIF section.

[nielsdos]

Is this an ICU change?

No idea.

In which case we need to duplicate the test and add some version constraint in the SKIPIF section.

We don't need to duplicate, we can use EXPECTF with %d for the string length I believe. EDIT: and a %something for the whitespace probably...

[cmb69]

We may need to backport 8dd51b4 to PHP 8.1 and 8.2.

[nielsdos]

Ah I didn't see that commit, it's exactly the issue mentioned in the commit message I'm encountering. Yeah it does sound like it needs a backport and that would fix it.