libicui18n fails to compile on Windows with MSVC 19.36.32502 due to P2468R2

[wroyca]

The meaning of == and != expressions has been altered in C++20 with the proposals P1185R2 and P2468R2, along with changes to how overload resolution is applied to them, resulting in previously valid code becoming invalid. As a result, ICU now produces error C2666 on certain overloaded functions and operators. Due to certain packages such as Qt being dependent on ICU, it also becomes infeasible to compile them.

    https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2022/p2468r2.html

b
c++ ..\echo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{fmtable} -> ..\echo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{fmtable}
c++ ..\echo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{fmtable} -> ..\echo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\obja{fmtable}
C:\Users\wroy\Desktop\echo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\fmtable.cpp(59): error C2666: 'icu_65::Measure::operator ==': overloaded functions have similar conversions
C:\Users\wroy\Desktop\echo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\unicode/measure.h(90): note: could be 'UBool icu_65::Measure::operator ==(const icu_65::UObject &) const'
C:\Users\wroy\Desktop\echo-msvc\libicuuc-65.1.0+6\libicu/uc\unicode/stringpiece.h(242): note: or       'UBool icu_65::operator ==(const icu_65::StringPiece &,const icu_65::StringPiece &)'
C:\Users\wroy\Desktop\echo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\unicode/measure.h(90): note: or 'UBool icu_65::Measure::operator ==(const icu_65::UObject &) const' [synthesized expression 'y == x']
C:\Users\wroy\Desktop\echo-msvc\libicuuc-65.1.0+6\libicu/uc\unicode/stringpiece.h(242): note: or 'UBool icu_65::operator ==(const icu_65::StringPiece &,const icu_65::StringPiece &)' [synthesized expression 'y == x']
C:\Users\wroy\Desktop\echo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\fmtable.cpp(59): note: while trying to match the argument list '(const icu_65::Measure, const icu_65::Measure)'
error: process cl exited with code 2

[boris-kolpackov]

Thanks for the report. Do you know by any chance if this has been fixed in newer versions of ICU? We are overdue for an upgrade.

[wroyca]

From what I understand, it seems like they've decided to use Wno-ambiguous-reversed-operator instead, but I wasn't able to see for myself as I had other things to focus on. However, I should have some free time on Monday to check it out.

Relevant discussions:

https://unicode-org.atlassian.net/browse/ICU-22014
https://unicode-org.atlassian.net/browse/ICU-20973

[wroyca]

Here is a git patch that addresses the compilation issue. Note that it has not undergone extensive testing.

0001-Fix-P2468R2-with-MSVC-19.36.32502.patch

From 3fbe14993f0e2c4a1c6f490b71ba56d682069060 Mon Sep 17 00:00:00 2001
From: = <wroy@proton.me>
Date: Mon, 20 Mar 2023 11:10:38 -0400
Subject: [PATCH] Fix P2468R2 with MSVC 19.36.32502

---
 icu4c/source/i18n/fmtable.cpp | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/icu4c/source/i18n/fmtable.cpp b/icu4c/source/i18n/fmtable.cpp
index dbfd3c26ba..33c0ad8505 100644
--- a/icu4c/source/i18n/fmtable.cpp
+++ b/icu4c/source/i18n/fmtable.cpp
@@ -56,7 +56,8 @@ using number::impl::DecimalQuantity;
 // Return TRUE if *a == *b.
 static inline UBool objectEquals(const UObject* a, const UObject* b) {
     // LATER: return *a == *b;
-    return *((const Measure*) a) == *((const Measure*) b);
+    return std::memcmp(static_cast<const Measure*>(a),
+                       static_cast<const Measure*>(b), sizeof(Measure)) == 0;
 }
 
 // Return a clone of *a.
-- 
2.40.0.windows.1

[boris-kolpackov]

No, I am 99% sure this patch is wrong (you are most likely comparing padding bytes inside Measure). I think it's best to just wait until upstream releases a fix (whatever it might be) and then we find time to release a new version. If you can ping this issue when that happens (that is, upstream releases a fixed version), that would be helpful.

[wroyca]

No, I am 99% sure this patch is wrong (you are most likely comparing padding bytes inside Measure).

I don't believe the compiler would insert padding for the measure object - Even if it did, the code should still be accurate since we are already comparing the entire type, aren't we?

*((any type ptr)param) == *((any type ptr)param)

In any case, I am hesitant to rely on "waiting for upstream to release a fix," especially given their decision to use the flag approach. It could take a considerable amount of time to resolve the issue, and in the meantime, we are unable to compile certain packages because they depend on the ICU package, which happens to be a single point of failure on Windows.

Should this patch cause issues as you have indicated, adding the overload to the Measure class itself would be the suitable solution - however, this could result in more significant consequences than the objectEquals method ever would

[wroyca]

@boris-kolpackov Here's the Patch that add the overload to the Measure class itself:

0001-Add-operator-for-P2468R2.patch

From f37efc2c29d25ac774890e272b7accd712448631 Mon Sep 17 00:00:00 2001
From: William Roy <wroy@proton.me>
Date: Tue, 21 Mar 2023 11:49:00 -0400
Subject: [PATCH] Add operator!= for P2468R2

---
 icu4c/source/i18n/unicode/measure.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/icu4c/source/i18n/unicode/measure.h b/icu4c/source/i18n/unicode/measure.h
index fa9c29351e..186751074a 100644
--- a/icu4c/source/i18n/unicode/measure.h
+++ b/icu4c/source/i18n/unicode/measure.h
@@ -88,6 +88,7 @@ class U_I18N_API Measure: public UObject {
      * @stable ICU 3.0
      */
     UBool operator==(const UObject& other) const;
+    UBool operator!=(const UObject& other) const;
 
     /**
      * Return a reference to the numeric value of this object.  The
-- 
2.40.0.windows.1

FYI we still need to create the definition in measure.cpp, e.g:

0001-Add-operator-definition-for-P2468R2.patch

From b07978eef67ba613d83c57fca3ec12ecebf599eb Mon Sep 17 00:00:00 2001
From: William Roy <wroy@proton.me>
Date: Tue, 21 Mar 2023 12:00:32 -0400
Subject: [PATCH] Add operator != definition for P2468R2

---
 icu4c/source/i18n/measure.cpp | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/icu4c/source/i18n/measure.cpp b/icu4c/source/i18n/measure.cpp
index bffa44215e..e73f92d81b 100644
--- a/icu4c/source/i18n/measure.cpp
+++ b/icu4c/source/i18n/measure.cpp
@@ -69,6 +69,10 @@ UBool Measure::operator==(const UObject& other) const {
         (unit == NULL || *unit == *m.unit);
 }
 
+UBool Measure::operator!=(const UObject& other) const {
+  return !(this->operator==(other));
+}
+
 U_NAMESPACE_END
 
 #endif // !UCONFIG_NO_FORMATTING
-- 
2.40.0.windows.1

[boris-kolpackov]

This looks better. Can you confirm that this fixes all the compilation errors in this MSVC version?

[wroyca]

This looks better. Can you confirm that this fixes all the compilation errors in this MSVC version?

It does, I use it currently👍

[boris-kolpackov]

@wroyca Do you know if there is an upstream's official fix/patch for this yet?

[boris-kolpackov]

Also, 19.36.32502 is (currently) the preview version of MSVC 17.6, correct?

[wroyca]

Also, 19.36.32502 is (currently) the preview version of MSVC 17.6, correct?

Sorry late answer (timezone) but that's correct, as for upstream I didn't have a chance to look again, but you can quickly check for the overload upstream, if it's missing then the issue is still present - I'll be home a bit later today to check again

[boris-kolpackov]

I just checked 72.1 and while they didn't add the != overload, they did change the return type to bool. I've also checked the current development version (the main branch) and there is nothing new there in this regard. So perhaps that's enough? It would have been helpful if you could test this change (i.e., change UBool to bool instead of adding the != overload) and confirm it helps (we don't deploy preview versions of MSVC to our CI so we cannot test this ourselves).

[wroyca]

I just checked 72.1 and while they didn't add the != overload, they did change the return type to bool. I've also checked the current development version (the main branch) and there is nothing new there in this regard. So perhaps that's enough? It would have been helpful if you could test this change (i.e., change UBool to bool instead of adding the != overload) and confirm it helps (we don't deploy preview versions of MSVC to our CI so we cannot test this ourselves).

Thank, I'll take a look and come back to you as soon as possible 👍

[wroyca]

@boris-kolpackov Confirmed that the issue is still present, but at a different location this time 👍

PS C:\Users\wroy\Desktop\foo> b
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{fmtable} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{fmtable}
C:\Users\wroy\Desktop\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\fmtable.cpp(59): error C2666: 'icu_65::Measure::operator ==': overloaded functions have similar conversions
C:\Users\wroy\Desktop\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\unicode/measure.h(90): note: could be 'UBool icu_65::Measure::operator ==(const icu_65::UObject &) const'
C:\Users\wroy\Desktop\foo-msvc\libicuuc-65.1.0+6\libicu/uc\unicode/stringpiece.h(242): note: or       'UBool icu_65::operator ==(const icu_65::StringPiece &,const icu_65::StringPiece &)'
C:\Users\wroy\Desktop\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\unicode/measure.h(90): note: or 'UBool icu_65::Measure::operator ==(const icu_65::UObject &) const' [synthesized expression 'y == x']
C:\Users\wroy\Desktop\foo-msvc\libicuuc-65.1.0+6\libicu/uc\unicode/stringpiece.h(242): note: or 'UBool icu_65::operator ==(const icu_65::StringPiece &,const icu_65::StringPiece &)' [synthesized expression 'y == x']
C:\Users\wroy\Desktop\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\fmtable.cpp(59): note: while trying to match the argument list '(const icu_65::Measure, const icu_65::Measure)'
error: process C:\Program Files\Microsoft Visual Studio\2022\Preview\VC\Tools\MSVC\14.36.32502\bin\Hostx64\x64\cl exited with code 2
  info: command line: "C:\Program Files\Microsoft Visual Studio\2022\Preview\VC\Tools\MSVC\14.36.32502\bin\Hostx64\x64\cl" -IC:\Users\wroy\Desktop\foo-msvc\libicui18n-65.1.0+6\libicui18n/i18n -D_CRT_SECURE_NO_WARNINGS -D_SCL_SECURE_NO_WARNINGS -DU_I18N_IMPLEMENTATION -DU_ATTRIBUTE_DEPRECATED= -DU_HAVE_STRTOD_L=0 -DHAVE_DLOPEN=0 -DU_HAVE_MMAP=0 -DWIN32 -DU_HAVE_DIRENT_H=0 -DU_HAVE_POPEN=0 -DU_HAVE_TZNAME=0 -IC:\Users\wroy\Desktop\foo-msvc\libicuuc-65.1.0+6\libicu/uc /wd4251 /wd4275 /wd4800 /utf-8 /Zc:wchar_t /GF /Gy /wd4244 /wd4996 /nologo /std:c++20 /Zc:__cplusplus /I "C:\Program Files\Microsoft Visual Studio\2022\Preview\VC\Tools\MSVC\14.36.32502\include" /I "C:\Program Files (x86)\Windows Kits\10\Include\10.0.22000.0\ucrt" /I "C:\Program Files (x86)\Windows Kits\10\Include\10.0.22000.0\shared" /I "C:\Program Files (x86)\Windows Kits\10\Include\10.0.22000.0\um" /I "C:\Program Files (x86)\Windows Kits\10\Include\10.0.22000.0\winrt" /utf-8 /EHsc /MD /Fo: ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\fmtable.dll.obj /c /TP C:\Users\wroy\Desktop\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\fmtable.cpp
  info: while updating ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{fmtable}
  info: while updating ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\libs{icui18n}
  info: while updating ..\foo-msvc\libQt6Core-6.4.3-a.0.20230329144408.6712fd545990\QtCore\libs{Qt6Core}
  info: while updating ..\foo-msvc\foo\foo\exe{foo}
  info: while updating ..\foo-msvc\foo\dir{foo\}
  info: while updating ..\foo-msvc\dir{foo\}

However, my patch is still applicable and resolve it:

0001-Add-operator-for-P2468R2.patch
0001-Add-operator-definition-for-P2468R2.patch

b
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{fmtable} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{fmtable}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{choicfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{choicfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{compactdecimalformat} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{compactdecimalformat}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{curramt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{curramt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{dtptngen} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{dtptngen}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{dcfmtsym} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{dcfmtsym}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{currfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{currfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{dtitvinf} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{dtitvinf}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{decimfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{decimfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{datefmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{datefmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{dtfmtsym} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{dtfmtsym}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{currpinf} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{currpinf}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{formattedval_sbimpl} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{formattedval_sbimpl}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{quantityformatter} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{quantityformatter}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{measure} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{measure}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{measfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{measfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{nfrs} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{nfrs}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{msgfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{msgfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{nfrule} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{nfrule}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{nfsubs} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{nfsubs}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_affixutils} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_affixutils}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_asformat} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_asformat}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_capi} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_capi}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_compact} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_compact}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_currencysymbols} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_currencysymbols}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_decimalquantity} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_decimalquantity}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_decimfmtprops} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_decimfmtprops}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_fluent} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_fluent}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_formatimpl} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_formatimpl}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_grouping} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_grouping}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_integerwidth} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_integerwidth}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{tmutamt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{tmutamt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_longnames} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_longnames}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_mapper} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_mapper}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_modifiers} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_modifiers}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_multiplier} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_multiplier}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_notation} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_notation}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_output} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_output}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_padding} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_padding}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_patternmodifier} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_patternmodifier}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_patternstring} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_patternstring}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_rounding} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_rounding}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_scientific} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_scientific}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{udat} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{udat}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_skeletons} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_skeletons}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{number_utils} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{number_utils}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{numfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{numfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{numparse_affixes} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{numparse_affixes}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{numparse_compositions} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{numparse_compositions}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{numparse_currency} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{numparse_currency}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{numparse_decimal} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{numparse_decimal}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{numparse_impl} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{numparse_impl}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{numparse_parsednumber} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{numparse_parsednumber}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{numparse_scientific} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{numparse_scientific}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{numparse_symbols} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{numparse_symbols}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{numparse_validators} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{numparse_validators}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{numrange_fluent} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{numrange_fluent}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{numrange_impl} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{numrange_impl}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{windtfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{windtfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{plurfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{plurfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{plurrule} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{plurrule}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{rbnf} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{rbnf}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{reldatefmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{reldatefmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{reldtfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{reldtfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{scientificnumberformatter} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{scientificnumberformatter}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{selfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{selfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{simpletz} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{simpletz}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{smpdtfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{smpdtfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{string_segment} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{string_segment}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{timezone} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{timezone}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{tmutfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{tmutfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{translit} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{translit}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{udateintervalformat} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{udateintervalformat}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\cxx{dtitvfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\objs{dtitvfmt}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{udatpg} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{udatpg}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{umsg} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{umsg}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{unum} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{unum}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{upluralrules} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{upluralrules}
c++ ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\cxx{winnmfmt} -> ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\i18n\objs{winnmfmt}
ld ..\foo-msvc\libicui18n-65.1.0+6\libicui18n\libs{icui18n}
ld ..\foo-msvc\libQt6Core-6.4.3-a.0.20230329144408.6712fd545990\QtCore\libs{Qt6Core}
ld ..\foo-msvc\libQt6Gui-6.4.3-a.0.20230329144408.6712fd545990\QtGui\libs{Qt6Gui}
ld ..\foo-msvc\libQt6Widgets-6.4.3-a.0.20230329144408.6712fd545990\QtWidgets\libs{Qt6Widgets}
ld ..\foo-msvc\foo\foo\exe{foo}
ln ..\foo-msvc\foo\foo\exe{foo} -> foo\
ld ..\foo-msvc\libQt6Gui-6.4.3-a.0.20230329144408.6712fd545990\QtGuiPlugins\imageformats\libs{qgif}
ld ..\foo-msvc\libQt6Gui-6.4.3-a.0.20230329144408.6712fd545990\QtGuiPlugins\imageformats\libs{qico}
ld ..\foo-msvc\libQt6Gui-6.4.3-a.0.20230329144408.6712fd545990\QtGuiPlugins\imageformats\libs{qjpeg}
ld ..\foo-msvc\libQt6Gui-6.4.3-a.0.20230329144408.6712fd545990\QtGuiPlugins\platforms\libs{qwindows}
ld ..\foo-msvc\libQt6Widgets-6.4.3-a.0.20230329144408.6712fd545990\QtWidgetsPlugins\styles\libs{qwindowsvistastyle}

[boris-kolpackov]

Confirmed that the issue is still present [...]

measure.h(90): note: could be 'UBool icu_65::Measure::operator ==(const icu_65::UObject &) const'

Not sure what this is confirming since it still seems to use UBool. Can you: (1) revert your patch (or rebuild from scratch), (2) replace Ubool with bool (both in declaration and definition) and see if that helps.