Do not inherit LC_CTYPE locale from environment

Treatment of locales in PHP is currently inconsistent: The LC_ALL
locale is set to "C", as is standard behavior on program startup.
The LC_CTYPE locale is set to "", which will inherit it from the
environment. However, the inherited LC_CTYPE locale will only be
used in some cases, while in other cases it is necessary to perform
an explicit setlocale() call in PHP first. This is the case for
the locale-sensitive handling in the PCRE extension.

Make things consistent by *never* inheriting any locales from the
environment. LC_ALL, including LC_CTYPE will be "C" on startup.
A locale can be set or inherited through an explicit setlocale()
call, at which point the behavior will be fully consistent and
predictable.

Closes GH-5488.

1 files changed, 6 insertions, 0 deletions

diff --git a/UPGRADING b/UPGRADING
index 96b9745d4c..982ebe635c 100644
--- a/UPGRADING
+++ b/UPGRADING
@@ -397,6 +397,12 @@ PHP 8.0 UPGRADE NOTES
   . If the array returned by __sleep() contains non-existing properties, these
     are now silently ignored. Previously, such properties would have been
     serialized as if they had the value NULL.
+  . The default locale on startup is now always "C". No locales are inherited
+    from the environment by default. Previously, LC_ALL was set to "C", while
+    LC_CTYPE was inherited from the environment. However, some functions did not
+    respect the inherited locale without an explicit setlocale() call. An
+    explicit setlocale() call is now always required if you wish to change any
+    locale component from the default.
 
 - tidy:
   . The $use_include_path parameter, which was not used internally, has been