Bug 828442 – www-client/chromium-97.0.4692.36: ../../third_party/icu/source/i18n/unicode/msgfmt.h:432:19: error: virtual function 'operator==' has a different return type ('UBool' (aka 'signed char')) than the function it overrides (which has return type 'bool')

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description Alex Xu (Hello71) 2021-12-06 01:49:37 UTC]

In file included from ../../base/i18n/message_formatter.cc:14:
../../third_party/icu/source/i18n/unicode/msgfmt.h:432:19: error: virtual function 'operator==' has a different return type ('UBool' (aka 'signed char')) than the function it overrides (which has return type 'bool')
    virtual UBool operator==(const Format& other) const;
            ~~~~~ ^
/usr/include/unicode/format.h:114:18: note: overridden virtual function is here
    virtual bool operator==(const Format& other) const = 0;
            ~~~~ ^
In file included from ../../base/i18n/message_formatter.cc:14:
../../third_party/icu/source/i18n/unicode/msgfmt.h:1091:23: error: virtual function 'operator==' has a different return type ('UBool' (aka 'signed char')) than the function it overrides (which has return type 'bool')
        virtual UBool operator==(const Format&) const;
                ~~~~~ ^
/usr/include/unicode/format.h:114:18: note: overridden virtual function is here
    virtual bool operator==(const Format& other) const = 0;
            ~~~~ ^
2 errors generated.

seems to compile and run ok with:

--- a/www-client/chromium/chromium-97.0.4692.36.ebuild
+++ b/www-client/chromium/chromium-97.0.4692.36.ebuild
@@ -499,7 +499,7 @@ src_prepare() {
                keeplibs+=( third_party/libxslt )
                keeplibs+=( third_party/openh264 )
                keeplibs+=( third_party/re2 )
-               if use system-icu; then
+               if ! use system-icu; then
                        keeplibs+=( third_party/icu )
                fi
        fi

[Comment 1 Stephan Hartmann (RETIRED) 2021-12-06 08:29:31 UTC]

Please attach emerge --info and full build.log to this bug. What version of ICU is installed on your system?

[Comment 2 Alex Xu (Hello71) 2021-12-07 17:14:45 UTC]

Created attachment 757637 [details]
emerge --info chromium

I don't have the failed build log anymore. emerge --info chromium attached. icu:

[binary   R    ] dev-libs/icu-70.1-1:0/70.1::gentoo  USE="-debug -doc -examples -static-libs" ABI_X86="32 (64) (-x32)" 0 KiB

what is the logic for the reversal of this USE (if use system, then keep icu), if not a typo?

[Comment 3 Stephan Hartmann (RETIRED) 2021-12-07 17:55:03 UTC]

Is it possible that you edited the ebuild locally, because it is ! use system-icu in the tree:
https://gitweb.gentoo.org/repo/gentoo.git/tree/www-client/chromium/chromium-97.0.4692.36.ebuild#n486

[Comment 4 Stephan Hartmann (RETIRED) 2021-12-07 17:56:09 UTC]

Scratch that. You have set CHROMIUM_FORCE_LIBCXX=yes?

[Comment 5 Alex Xu (Hello71) 2021-12-07 18:07:31 UTC]

(In reply to Stephan Hartmann from comment #4)
> Scratch that. You have set CHROMIUM_FORCE_LIBCXX=yes?

yes, it is in the emerge --info. I assume it works ok with CHROMIUM_FORCE_LIBCXX=no, on the basis that nobody else seems to have complained, but I haven't tested it. I guess it would probably be broken again if CHROMIUM_FORCE_LIBCXX=yes is set as default in future, and the user had incompatible icu installed.

[Comment 6 Stephan Hartmann (RETIRED) 2021-12-07 18:19:18 UTC]

The logic in the ebuild is correct. With libcxx you cannot link against system ICU, because that might link against libstc++ from GCC. The unbundling script has problems if you specify library pathes twice. So we only add icu to keeplibs in case USE=system-icu.
The problem here seems to be that the bundled ICU includes headers from system ICU. We need the compiler command to debug the issue further.

[Comment 7 Alex Xu (Hello71) 2021-12-07 20:42:47 UTC]

ah, i see, it's backwards because it was potentially already kept previously; it has the function of [[ " ${keeplibs[*]} " == *" third_party/icu "* ]] || keeplibs+=( third_party/icu ). shouldn't https://gitweb.gentoo.org/repo/gentoo.git/tree/www-client/chromium/chromium-97.0.4692.36.ebuild#n603 be moved inside the following if block then?

[Comment 8 Krzysztof Nowicki 2022-01-23 20:14:39 UTC]

(In reply to Stephan Hartmann from comment #6)
> We need the compiler command to debug the issue further.

# clang++ -v -MMD -MF obj/base/i18n/time_formatting.o.d -DBASE_I18N_IMPLEMENTATION -DUSE_UDEV -DUSE_AURA=1 -DUSE_GLIB=1 -DUSE_NSS_CERTS=1 -DUSE_OZONE=1 -DUSE_X11=1 -DOFFICIAL_BUILD -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SO
URCE -D_LARGEFILE64_SOURCE -DNO_UNWIND_TABLES -D_GNU_SOURCE -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -D_LIBCPP_ABI_UNSTABLE -D_LIBCPP_DISABLE_VISIBILITY_ANNOTATIONS -D_LIBCXXABI_DISABLE_VISIBILITY_ANNOTATIONS -D_LIBCPP_ENABLE_NODISCARD -DCR_LIBCXX_REVISION=79a2e924d96e2fc1e4b937c42efd08898fa472d7 -DNDEBUG -DNV
ALGRIND -DDYNAMIC_ANNOTATIONS_ENABLED=0 -DGLIB_VERSION_MAX_ALLOWED=GLIB_VERSION_2_40 -DGLIB_VERSION_MIN_REQUIRED=GLIB_VERSION_2_40 -DUSING_SYSTEM_ICU=1 -DICU_UTIL_DATA_IMPL=ICU_UTIL_DATA_STATIC -DU_IMPORT=U_EXPORT -I../.. -Igen -I../../buildtools/third_party/libc++ -Igen/shim_headers/icui18n_shim -Igen/shim_headers/i
cuuc_shim -I../../third_party/perfetto/include -Igen/third_party/perfetto/build_config -Igen/third_party/perfetto -Igen/shim_headers/zlib_shim -I../../third_party/ced/src -I../../third_party/abseil-cpp -I../../third_party/boringssl/src/include -I../../third_party/protobuf/src -Igen/protoc_out -fno-delete-null-pointer
-checks -fno-ident -fno-strict-aliasing --param=ssp-buffer-size=4 -fstack-protector -fno-unwind-tables -fno-asynchronous-unwind-tables -fPIC -pthread -fcolor-diagnostics -fmerge-all-constants -flto=thin -fsplit-lto-unit -fwhole-program-vtables -fcomplete-member-pointers -no-canonical-prefixes -Wimplicit-fallthrough -
Wunreachable-code-aggressive -Wthread-safety -Wextra-semi -Wno-missing-field-initializers -Wno-unused-parameter -Wloop-analysis -Wno-unneeded-internal-declaration -Wenum-compare-conditional -Wno-psabi -Wno-ignored-pragma-optimize -fno-omit-frame-pointer -ftrivial-auto-var-init=pattern -fprofile-instr-use=../../chrome
/build/pgo_profiles/chrome-linux-4692-1640701189-f21838a98aa9a4bd13520a4261d2d77cbab9c3f3.profdata -Wno-profile-instr-unprofiled -Wno-profile-instr-out-of-date -Wno-backend-plugin -fvisibility=hidden -Wheader-hygiene -Wstring-conversion -Wtautological-overlap-compare -Wexit-time-destructors -I/usr/include/glib-2.0 -I
/usr/lib64/glib-2.0/include -I/usr/lib64/libffi/include -DPROTOBUF_ALLOW_DEPRECATED=1 -std=c++14 -fno-trigraphs -Wno-trigraphs -fno-exceptions -fno-rtti -nostdinc++ -isystem../../buildtools/third_party/libc++/trunk/include -isystem../../buildtools/third_party/libc++abi/trunk/include -fvisibility-inlines-hidden -O2 -p
ipe -march=haswell -fno-ident -msse4.2 -mavx -fno-semantic-interposition -falign-functions=32 -flto=thin -stdlib=libc++ -Wno-unknown-warning-option -c ../../base/i18n/time_formatting.cc -o obj/base/i18n/time_formatting.o
clang version 13.0.0
Target: x86_64-pc-linux-gnu
Thread model: posix
InstalledDir: /usr/lib/llvm/13/bin
Selected GCC installation: /usr/lib/gcc/x86_64-pc-linux-gnu/11.2.0
Candidate multilib: .;@m64
Candidate multilib: 32;@m32
Selected multilib: .;@m64
clang++: warning: argument unused during compilation: '-stdlib=libc++' [-Wunused-command-line-argument]
 (in-process)
 "/usr/lib/llvm/13/bin/clang++" -cc1 -triple x86_64-pc-linux-gnu -emit-llvm-bc -flto=thin -flto-unit -disable-free -disable-llvm-verifier -discard-value-names -main-file-name time_formatting.cc -function-alignment 5 -mrelocation-model pic -pic-level 2 -fmerge-all-constants -fno-delete-null-pointer-checks -mframe-poin
ter=all -relaxed-aliasing -fmath-errno -fno-rounding-math -mconstructor-aliases -target-cpu haswell -target-feature +sse4.2 -target-feature +avx -mllvm -treat-scalable-fixed-error-as-warning -debugger-tuning=gdb -v -fprofile-instrument-use-path=../../chrome/build/pgo_profiles/chrome-linux-4692-1640701189-f21838a98aa9
a4bd13520a4261d2d77cbab9c3f3.profdata -fcoverage-compilation-dir=/var/tmp/portage/www-client/chromium-97.0.4692.71-r1/work/chromium-97.0.4692.71/out/Release -nostdinc++ -resource-dir /usr/lib/llvm/13/bin/../../../../lib/clang/13.0.0 -dependency-file obj/base/i18n/time_formatting.o.d -MT obj/base/i18n/time_formatting.
o -isystem ../../buildtools/third_party/libc++/trunk/include -isystem ../../buildtools/third_party/libc++abi/trunk/include -D BASE_I18N_IMPLEMENTATION -D USE_UDEV -D USE_AURA=1 -D USE_GLIB=1 -D USE_NSS_CERTS=1 -D USE_OZONE=1 -D USE_X11=1 -D OFFICIAL_BUILD -D _FILE_OFFSET_BITS=64 -D _LARGEFILE_SOURCE -D _LARGEFILE64_S
OURCE -D NO_UNWIND_TABLES -D _GNU_SOURCE -D __STDC_CONSTANT_MACROS -D __STDC_FORMAT_MACROS -D _LIBCPP_ABI_UNSTABLE -D _LIBCPP_DISABLE_VISIBILITY_ANNOTATIONS -D _LIBCXXABI_DISABLE_VISIBILITY_ANNOTATIONS -D _LIBCPP_ENABLE_NODISCARD -D CR_LIBCXX_REVISION=79a2e924d96e2fc1e4b937c42efd08898fa472d7 -D NDEBUG -D NVALGRIND -D
 DYNAMIC_ANNOTATIONS_ENABLED=0 -D GLIB_VERSION_MAX_ALLOWED=GLIB_VERSION_2_40 -D GLIB_VERSION_MIN_REQUIRED=GLIB_VERSION_2_40 -D USING_SYSTEM_ICU=1 -D ICU_UTIL_DATA_IMPL=ICU_UTIL_DATA_STATIC -D U_IMPORT=U_EXPORT -I ../.. -I gen -I ../../buildtools/third_party/libc++ -I gen/shim_headers/icui18n_shim -I gen/shim_headers/
icuuc_shim -I ../../third_party/perfetto/include -I gen/third_party/perfetto/build_config -I gen/third_party/perfetto -I gen/shim_headers/zlib_shim -I ../../third_party/ced/src -I ../../third_party/abseil-cpp -I ../../third_party/boringssl/src/include -I ../../third_party/protobuf/src -I gen/protoc_out -I /usr/includ
e/glib-2.0 -I /usr/lib64/glib-2.0/include -I /usr/lib64/libffi/include -D PROTOBUF_ALLOW_DEPRECATED=1 -internal-isystem /usr/lib/llvm/13/bin/../../../../lib/clang/13.0.0/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-pc-linux-gnu/11.2.0/../../../../x86_64-pc-linux-gnu/include -inte
rnal-externc-isystem /include -internal-externc-isystem /usr/include -O2 -Wimplicit-fallthrough -Wunreachable-code-aggressive -Wthread-safety -Wextra-semi -Wno-missing-field-initializers -Wno-unused-parameter -Wloop-analysis -Wno-unneeded-internal-declaration -Wenum-compare-conditional -Wno-psabi -Wno-ignored-pragma-
optimize -Wno-profile-instr-unprofiled -Wno-profile-instr-out-of-date -Wno-backend-plugin -Wheader-hygiene -Wstring-conversion -Wtautological-overlap-compare -Wexit-time-destructors -Wno-trigraphs -Wno-unknown-warning-option -std=c++14 -fno-trigraphs -fdeprecated-macro -fdebug-compilation-dir=/var/tmp/portage/www-cli
ent/chromium-97.0.4692.71-r1/work/chromium-97.0.4692.71/out/Release -ferror-limit 19 -fvisibility hidden -fvisibility-inlines-hidden -pthread -stack-protector 1 -stack-protector-buffer-size 4 -ftrivial-auto-var-init=pattern -fno-rtti -fgnuc-version=4.2.1 -Qn -fcolor-diagnostics -vectorize-loops -vectorize-slp -fwhole
-program-vtables -fsplit-lto-unit -fcomplete-member-pointers -o obj/base/i18n/time_formatting.o -x c++ ../../base/i18n/time_formatting.cc
clang -cc1 version 13.0.0 based upon LLVM 13.0.0 default target x86_64-pc-linux-gnu
ignoring nonexistent directory "gen/protoc_out"
ignoring nonexistent directory "/usr/local/include"
ignoring nonexistent directory "/usr/lib/gcc/x86_64-pc-linux-gnu/11.2.0/../../../../x86_64-pc-linux-gnu/include"
ignoring nonexistent directory "/include"
#include "..." search starts here:
#include <...> search starts here:
 ../..
 gen
 ../../buildtools/third_party/libc++
 gen/shim_headers/icui18n_shim
 gen/shim_headers/icuuc_shim
 ../../third_party/perfetto/include
 gen/third_party/perfetto/build_config
 gen/third_party/perfetto
 gen/shim_headers/zlib_shim
 ../../third_party/ced/src
 ../../third_party/abseil-cpp
 ../../third_party/boringssl/src/include
 ../../third_party/protobuf/src
 /usr/include/glib-2.0
 /usr/lib64/glib-2.0/include
 /usr/lib64/libffi/include
 ../../buildtools/third_party/libc++/trunk/include
 ../../buildtools/third_party/libc++abi/trunk/include
 /usr/lib/llvm/13/bin/../../../../lib/clang/13.0.0/include
 /usr/include
End of search list.
In file included from ../../base/i18n/time_formatting.cc:19:
../../third_party/icu/source/i18n/unicode/datefmt.h:238:19: error: virtual function 'operator==' has a different return type ('UBool' (aka 'signed char')) than the function it overrides (which has return type 'bool')
    virtual UBool operator==(const Format&) const;
            ~~~~~ ^
/usr/include/unicode/format.h:114:18: note: overridden virtual function is here
    virtual bool operator==(const Format& other) const = 0;
            ~~~~ ^
In file included from ../../base/i18n/time_formatting.cc:20:
In file included from ../../third_party/icu/source/i18n/unicode/dtitvfmt.h:29:
/usr/include/unicode/smpdtfmt.h:879:18: error: virtual function 'operator==' has a different return type ('bool') than the function it overrides (which has return type 'UBool' (aka 'signed char'))
    virtual bool operator==(const Format& other) const override;
            ~~~~ ^
../../third_party/icu/source/i18n/unicode/datefmt.h:238:19: note: overridden virtual function is here
    virtual UBool operator==(const Format&) const;
            ~~~~~ ^
In file included from ../../base/i18n/time_formatting.cc:20:
../../third_party/icu/source/i18n/unicode/dtitvfmt.h:446:19: error: virtual function 'operator==' has a different return type ('UBool' (aka 'signed char')) than the function it overrides (which has return type 'bool')
    virtual UBool operator==(const Format& other) const;
            ~~~~~ ^
/usr/include/unicode/format.h:114:18: note: overridden virtual function is here
    virtual bool operator==(const Format& other) const = 0;
            ~~~~ ^
In file included from ../../base/i18n/time_formatting.cc:23:
../../third_party/icu/source/i18n/unicode/measfmt.h:151:19: error: virtual function 'operator==' has a different return type ('UBool' (aka 'signed char')) than the function it overrides (which has return type 'bool')
    virtual UBool operator==(const Format &other) const;
            ~~~~~ ^
/usr/include/unicode/format.h:114:18: note: overridden virtual function is here
    virtual bool operator==(const Format& other) const = 0;
            ~~~~ ^
4 errors generated.

[Comment 9 Krzysztof Nowicki 2022-01-24 09:17:12 UTC]

(In reply to Stephan Hartmann from comment #6)
> The logic in the ebuild is correct. With libcxx you cannot link against
> system ICU, because that might link against libstc++ from GCC. The
> unbundling script has problems if you specify library pathes twice. So we
> only add icu to keeplibs in case USE=system-icu.
> The problem here seems to be that the bundled ICU includes headers from
> system ICU. We need the compiler command to debug the issue further.

I have tried to disabling "system-icu" and Chromium built fine witg clang and libc++. This means that the logic in the ebuild is not entirely correct.

There are three changes in the build logic that are done differently dependent of the "system-icu" flag:
 1. keeplibs - used to control the unbubdler
 2. setting of gn_system_libraries
 3. setting of icu_use_data_file=false

Only the first action is reverted in presence of CHROMIUM_FORCE_LIBCXX=true. The remaining two steps are executed even with "+system-icu" and libc++.

From my naive point of view it seems that the general behaviour should be to block use of "system-icu" together with CHROMIUM_FORCE_LIBCXX. This means that the ebuild should either:
 a) fix steps 2 and 3 to also behave differently when libc++ is enabled
 b) throw an error if "system-icu" is requested together with enabling libc++.

Personally I would vote for b).

A nice addition would actually be to convert CHROMIUM_FORCE_CLANG and CHROMIUM_FORCE_LIBCXX to USE flags, which would allow to set-up REQUIRED_USE to block unsupported set of options.

[Comment 10 Stephan Hartmann (RETIRED) 2022-01-24 09:47:24 UTC]

I have added USE=libcxx in ebuilds in my overlay, e.g.:

https://github.com/stha09/gpo-stha09/blob/master/www-client/chromium/chromium-99.0.4840.0.ebuild

However, there are a lot more changes (USE=headless, gtk4, cross-compiling), but I think those are not ready for the tree.

[Comment 11 Larry the Git Cow 2022-01-30 08:37:43 UTC]

The bug has been referenced in the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=c65f1fb18943313a621d8cce43d0a68ebf65d2f4

commit c65f1fb18943313a621d8cce43d0a68ebf65d2f4
Author:     Stephan Hartmann <[e-mail]>
AuthorDate: 2022-01-30 08:36:59 +0000
Commit:     Stephan Hartmann <[e-mail]>
CommitDate: 2022-01-30 08:37:35 +0000

    www-client/chromium: dev channel bump to 99.0.4844.11
    
    - improve headless builds
    - fix cross compiling
    - remove CHROMIUM_FORCE_LIBCXX and add USE=libcxx instead
    - allow building with GTK4
    
    Bug: https://bugs.gentoo.org/786597
    Bug: https://bugs.gentoo.org/828442
    Package-Manager: Portage-3.0.28, Repoman-3.0.3
    Signed-off-by: Stephan Hartmann <[e-mail]>

 www-client/chromium/Manifest                       |   4 +-
 ....4840.0.ebuild => chromium-99.0.4844.11.ebuild} | 242 +++++++++++++--------
 .../chromium/files/chromium-cross-compile.patch    |  12 +
 www-client/chromium/metadata.xml                   |   2 +
 4 files changed, 162 insertions(+), 98 deletions(-)

[Comment 12 Alex Xu (Hello71) 2022-01-31 21:05:38 UTC]

this seems like a good change, but am i missing something or is the icu problem not fixed here? there is no special handling of USE=libcxx && USE=system-icu. is it supposed to have REQUIRED_USE="libcxx? ( !system-icu )"?

[Comment 13 Stephan Hartmann (RETIRED) 2022-02-01 11:40:14 UTC]

(In reply to Alex Xu (Hello71) from comment #12)
> this seems like a good change, but am i missing something or is the icu
> problem not fixed here? there is no special handling of USE=libcxx &&
> USE=system-icu. is it supposed to have REQUIRED_USE="libcxx? ( !system-icu
> )"?

In the previous configuration we were forced to use bundled libraries that link against libstdc++. With USE=libcxx we are using bundled libcxx and link static with it. Only re2 uses STL containers in its ABI and we can unbundle other libraries including ICU again.

[Comment 14 Alex Xu (Hello71) 2022-02-01 14:01:51 UTC]

(In reply to Stephan Hartmann from comment #13)
> (In reply to Alex Xu (Hello71) from comment #12)
> > this seems like a good change, but am i missing something or is the icu
> > problem not fixed here? there is no special handling of USE=libcxx &&
> > USE=system-icu. is it supposed to have REQUIRED_USE="libcxx? ( !system-icu
> > )"?
> 
> In the previous configuration we were forced to use bundled libraries that
> link against libstdc++. With USE=libcxx we are using bundled libcxx and link
> static with it. Only re2 uses STL containers in its ABI and we can unbundle
> other libraries including ICU again.

ah, I see. that makes sense. it seems like it should work, but i can't test this now; i'm happy to close this with TEST-REQUEST or leave open, whichever you prefer.