unplugged-system/external/cronet/third_party/icu/patches/wordbrk.patch

diff --git a/source/data/brkitr/rules/word.txt b/source/data/brkitr/rules/word.txt
index 0f0e734d27c0..b46038237859 100644
--- a/source/data/brkitr/rules/word.txt
+++ b/source/data/brkitr/rules/word.txt
@@ -38,12 +38,34 @@ $Regional_Indicator = [\p{Word_Break = Regional_Indicator}];
 $Format             = [\p{Word_Break = Format}];
 $Katakana           = [\p{Word_Break = Katakana}];
 $Hebrew_Letter      = [\p{Word_Break = Hebrew_Letter}];
-$ALetter            = [\p{Word_Break = ALetter} @];
+
+# Exclude '@' (commercial at, \u0040) from ALetter to maintain breaking at '@'.
+# ICU 49d192fefe09, in ICU 72, stopped breaking at '@' in order to not break up
+# e-mail addresses (https://unicode-org.atlassian.net/browse/CLDR-15767). In
+# light of the Chromium-specific change below that breaks on full-stop (period,
+# dot, \u002e below in MidNumLet), e-mail addresses will be broken in any case.
+# Thus, although the upstream intent was to not break "user.name@example.com" at
+# all, it actually would break down into {"user", ".", "name@example", ".",
+# "com"}, which is undesirable. See https://crbug.com/1410331. Maintain the
+# previous Chromium behavior of breaking at both '@' and '.'.
+#
+# TODO: Determine whether it's feasible to drop the Chromium-specific behaviors
+# (and thus this patch) for '.' and now '@'.
+$ALetter            = [\p{Word_Break = ALetter}];
+
 $Single_Quote       = [\p{Word_Break = Single_Quote}];
 $Double_Quote       = [\p{Word_Break = Double_Quote}];
-$MidNumLet          = [\p{Word_Break = MidNumLet}];
+
+# Remove two full stop characters from $MidNumLet and add them to $MidNum
+# to break a hostname into its components at the cost of breaking
+# 'e.g.' and 'i.e.' as well.
+# $MidNumLet is used in rules 6/7 (rules of our interest) and rules 11/12.
+# Because it's OR'd with $MidNum in rules 11/12, rules 11/12 are not affected
+# while rules 6/7 are reverted to the old behavior we want.
+$MidNumLet          = [[\p{Word_Break = MidNumLet}] - [\u002E \uFF0E]];
 $MidLetter          = [\p{Word_Break = MidLetter} - [\: \uFE55 \uFF1A]];
-$MidNum             = [\p{Word_Break = MidNum}];
+$MidNum             = [\p{Word_Break = MidNum}[\u002E \uFF0E]];
+
 $Numeric            = [\p{Word_Break = Numeric}];
 $ExtendNumLet       = [\p{Word_Break = ExtendNumLet}];
 $WSegSpace          = [\p{Word_Break = WSegSpace}];
diff --git a/source/data/brkitr/rules/word_POSIX.txt b/source/data/brkitr/rules/word_POSIX.txt
index e62fd7fac3b1..1115b464d835 100644
--- a/source/data/brkitr/rules/word_POSIX.txt
+++ b/source/data/brkitr/rules/word_POSIX.txt
@@ -38,12 +38,30 @@ $Regional_Indicator = [\p{Word_Break = Regional_Indicator}];
 $Format             = [\p{Word_Break = Format}];
 $Katakana           = [\p{Word_Break = Katakana}];
 $Hebrew_Letter      = [\p{Word_Break = Hebrew_Letter}];
-$ALetter            = [\p{Word_Break = ALetter} @];
+
+# Exclude '@' (commercial at, \u0040) from ALetter to maintain breaking at '@'.
+# ICU 49d192fefe09, in ICU 72, stopped breaking at '@' in order to not break up
+# e-mail addresses (https://unicode-org.atlassian.net/browse/CLDR-15767). In
+# light of the Chromium-specific change below that breaks on full-stop (period,
+# dot, \u002e below in MidNumLet), e-mail addresses will be broken in any case.
+# Thus, although the upstream intent was to not break "user.name@example.com" at
+# all, it actually would break down into {"user", ".", "name@example", ".",
+# "com"}, which is undesirable. See https://crbug.com/1410331. Maintain the
+# previous Chromium behavior of breaking at both '@' and '.'.
+#
+# TODO: Determine whether it's feasible to drop the Chromium-specific behaviors
+# (and thus this patch) for '.' and now '@'.
+$ALetter            = [\p{Word_Break = ALetter}];
+
 $Single_Quote       = [\p{Word_Break = Single_Quote}];
 $Double_Quote       = [\p{Word_Break = Double_Quote}];
-$MidNumLet          = [\p{Word_Break = MidNumLet} - [.]];
+
+# Remove full-width full stop (\uff0e) from $MidNumLet and add it to $MidNum, in
+# addition to the ordinary full stop (dot, period, '.', \u002e).
+$MidNumLet          = [[\p{Word_Break = MidNumLet}] - [\u002E \uFF0E]];
 $MidLetter          = [\p{Word_Break = MidLetter} - [\: \uFE55 \uFF1A]];
-$MidNum             = [\p{Word_Break = MidNum} [.]];
+$MidNum             = [\p{Word_Break = MidNum}[\u002E \uFF0E]];
+
 $Numeric            = [\p{Word_Break = Numeric}];
 $ExtendNumLet       = [\p{Word_Break = ExtendNumLet}];
 $WSegSpace          = [\p{Word_Break = WSegSpace}];
diff --git a/source/data/brkitr/rules/word_fi_sv.txt b/source/data/brkitr/rules/word_fi_sv.txt
index 544558f91adf..da79385318d0 100644
--- a/source/data/brkitr/rules/word_fi_sv.txt
+++ b/source/data/brkitr/rules/word_fi_sv.txt
@@ -38,12 +38,34 @@ $Regional_Indicator = [\p{Word_Break = Regional_Indicator}];
 $Format             = [\p{Word_Break = Format}];
 $Katakana           = [\p{Word_Break = Katakana}];
 $Hebrew_Letter      = [\p{Word_Break = Hebrew_Letter}];
-$ALetter            = [\p{Word_Break = ALetter} @];
+
+# Exclude '@' (commercial at, \u0040) from ALetter to maintain breaking at '@'.
+# ICU 49d192fefe09, in ICU 72, stopped breaking at '@' in order to not break up
+# e-mail addresses (https://unicode-org.atlassian.net/browse/CLDR-15767). In
+# light of the Chromium-specific change below that breaks on full-stop (period,
+# dot, \u002e below in MidNumLet), e-mail addresses will be broken in any case.
+# Thus, although the upstream intent was to not break "user.name@example.com" at
+# all, it actually would break down into {"user", ".", "name@example", ".",
+# "com"}, which is undesirable. See https://crbug.com/1410331. Maintain the
+# previous Chromium behavior of breaking at both '@' and '.'.
+#
+# TODO: Determine whether it's feasible to drop the Chromium-specific behaviors
+# (and thus this patch) for '.' and now '@'.
+$ALetter            = [\p{Word_Break = ALetter}];
+
 $Single_Quote       = [\p{Word_Break = Single_Quote}];
 $Double_Quote       = [\p{Word_Break = Double_Quote}];
-$MidNumLet          = [\p{Word_Break = MidNumLet}];
+
+# Remove two full stop characters from $MidNumLet and add them to $MidNum
+# to break a hostname into its components at the cost of breaking
+# 'e.g.' and 'i.e.' as well.
+# $MidNumLet is used in rules 6/7 (rules of our interest) and rules 11/12.
+# Because it's OR'd with $MidNum in rules 11/12, rules 11/12 are not affected
+# while rules 6/7 are reverted to the old behavior we want.
+$MidNumLet          = [[\p{Word_Break = MidNumLet}] - [\u002E \uFF0E]];
 $MidLetter          = [\p{Word_Break = MidLetter}];
-$MidNum             = [\p{Word_Break = MidNum}];
+$MidNum             = [\p{Word_Break = MidNum}[\u002E \uFF0E]];
+
 $Numeric            = [\p{Word_Break = Numeric}];
 $ExtendNumLet       = [\p{Word_Break = ExtendNumLet}];
 $WSegSpace          = [\p{Word_Break = WSegSpace}];