# Cpp2 exceptions **URL:** \ **Category:** General\ **Created:** [December 20, 2022, 1:06pm UTC](https://swi-prolog.discourse.group/t/cpp2-exceptions/6040 "2022-12-20T13:06:06Z")\ **Posts on this page:** 8\ **Page:** 5
**Author:** ![mgondan1](https://yyz2.discourse-cdn.com/free1/user_avatar/swi-prolog.discourse.group/mgondan1/32/803_2.png) [@mgondan1](https://swi-prolog.discourse.group/u/mgondan1)\ **Post date:** [December 27, 2025, 5:51pm UTC](https://swi-prolog.discourse.group/t/cpp2-exceptions/6040/85 "2025-12-27T17:51:18Z")
As far as I understand the unit tests, it is expected that PL\_get\_nchars fails in some of them (and raises a representation error). ```prolog test(hello, error(representation_error(encoding))) :- hello("世界", _Out). ``` ---
**Author:** ![peter.ludemann](https://yyz2.discourse-cdn.com/free1/user_avatar/swi-prolog.discourse.group/peter.ludemann/32/48_2.png) [@peter.ludemann](https://swi-prolog.discourse.group/u/peter.ludemann)\ **Post date:** [December 29, 2025, 7:08am UTC](https://swi-prolog.discourse.group/t/cpp2-exceptions/6040/86 "2025-12-29T07:08:01Z")
Yes, that test is designed to get an error. because “世界” can’t be represented in ASCII. Except, I can’t reproduce the error (see [clang.asan: unit tests fail · Issue #102 · SWI-Prolog/packages-cpp · GitHub](https://github.com/SWI-Prolog/packages-cpp/issues/102)) but instead shows up in `clib:crypt` for gcc-12, clang-14, but not clang-19. (I used `ctest j=1` to ensure the tests were isolated from each other, although that probably wasn’t necessary). `PlTerm::get_nchars()` uses `PlEx_fail()` to check whether the “false” return code from `PL_get_nchars()` is simply a failure or is an error (from reading the code, it seems that only a representation error is possible, but there might be other error possibilities). ---
**Author:** ![mgondan1](https://yyz2.discourse-cdn.com/free1/user_avatar/swi-prolog.discourse.group/mgondan1/32/803_2.png) [@mgondan1](https://swi-prolog.discourse.group/u/mgondan1)\ **Post date:** [December 29, 2025, 8:54pm UTC](https://swi-prolog.discourse.group/t/cpp2-exceptions/6040/87 "2025-12-29T20:54:42Z")
Unsure if this helps (maybe on the longer run), but I can reproduce this problem also in a minimalistic example: example.cpp: ```prolog #include PREDICATE(helloex, 0) { throw PlUnknownError("an exception"); } ``` Compile with `src/swipl-ld -g -shared example.cpp -o example.so`, then `use_foreign_library(example). helloex.` (cmake configuration as above) ```prolog CC="clang-21 -fsanitize=address,undefined -fno-sanitize=float-divide-by-zero -fno-sanitize=alignment -fno-sanitize=function -fno-omit-frame-pointer -std=gnu23" CXX="clang++-21 -fsanitize=address,undefined -fno-sanitize=float-divide-by-zero -fno-sanitize=alignment -fno-omit-frame-pointer -frtti -std=gnu++17" CFLAGS="-g -O3 -Wall " LDFLAGS="-Wl,-O1" CXXFLAGS="-g -O3 -Wall -pedantic -fpic" cmake -DCMAKE_CXX_STANDARD=17 -DCMAKE_INSTALL_PREFIX=~/swipl -DMACOSX_DEPENDENCIES_FROM=Homebrew -DSWIPL_INSTALL_IN_LIB=ON -DINSTALL_DOCUMENTATION=OFF -DBUILD_TESTING=ON -DINSTALL_TESTS=ON -DSWIPL_PACKAGES_JAVA=OFF -DSWIPL_PACKAGES_GUI=OFF -DSWIPL_PACKAGES_ODBC=OFF -DSWIPL_PACKAGES_PYTHON=OFF -S .. make ``` ---
**Author:** ![peter.ludemann](https://yyz2.discourse-cdn.com/free1/user_avatar/swi-prolog.discourse.group/peter.ludemann/32/48_2.png) [@peter.ludemann](https://swi-prolog.discourse.group/u/peter.ludemann)\ **Post date:** [December 30, 2025, 12:15am UTC](https://swi-prolog.discourse.group/t/cpp2-exceptions/6040/88 "2025-12-30T00:15:32Z")
> [@mgondan1](#): > > (cmake configuration as above) Did you set this up with `scripts/configure`? If so, what is the build directory name (with the options)? I’ll add this trivial test to the `cpp` tests. ---
**Author:** ![mgondan1](https://yyz2.discourse-cdn.com/free1/user_avatar/swi-prolog.discourse.group/mgondan1/32/803_2.png) [@mgondan1](https://swi-prolog.discourse.group/u/mgondan1)\ **Post date:** [December 30, 2025, 8:36am UTC](https://swi-prolog.discourse.group/t/cpp2-exceptions/6040/89 "2025-12-30T08:36:40Z")
No, with the cmake command shown. I’ll try to simplify this during the day, so that it matches the build.make.clang etc logic ---
**Author:** ![mgondan1](https://yyz2.discourse-cdn.com/free1/user_avatar/swi-prolog.discourse.group/mgondan1/32/803_2.png) [@mgondan1](https://swi-prolog.discourse.group/u/mgondan1)\ **Post date:** [December 30, 2025, 2:11pm UTC](https://swi-prolog.discourse.group/t/cpp2-exceptions/6040/90 "2025-12-30T14:11:27Z")
Simpler version: ```prolog cd swipl-devel mkdir build.make.clang\=21.asan cd build.make.clang\=21.asan ../scripts/configure make src/swipl-ld -g -shared example.cpp -o example.so src/swipl ?- use_foreign_library(example). ?- helloex. ``` Needs clang-21, of course. It can also be reproduced with with gcc (15.2.0 on my computer): ```prolog cd swipl-devel mkdir build.make.gcc.asan cd build.make.gcc.asan ../scripts/configure make src/swipl-ld -g -shared example.cpp -o example.so src/swipl ?- use_foreign_library(example). ?- helloex. ``` ---
**Author:** ![mgondan1](https://yyz2.discourse-cdn.com/free1/user_avatar/swi-prolog.discourse.group/mgondan1/32/803_2.png) [@mgondan1](https://swi-prolog.discourse.group/u/mgondan1)\ **Post date:** [January 3, 2026, 4:36pm UTC](https://swi-prolog.discourse.group/t/cpp2-exceptions/6040/91 "2026-01-03T16:36:24Z")
It seems that I can, well, “fix” the problem with ```prolog LD_PRELOAD="$(realpath "$(gcc -print-file-name=libasan.so)") $(realpath "$(gcc -print-file-name=libstdc++.so)")" ctest ``` so that it is rather a linker problem. See discussion of a similar problem here. [AddressSanitizer CHECK failed: ../../../sanitizer/asan/asan\_interceptors.cc:384 "((\_\_interception::real\_\_\_cxa\_throw)) when using --as-needed and dynamically loaded .so · Issue #934 · google/sanitizers · GitHub](https://github.com/google/sanitizers/issues/934) Needless to say that I have no idea what all this stuff is about. Edit: ```prolog LD_PRELOAD=$(gcc -print-file-name=libstdc++.so) ctest ``` works, too, if we add -DCMAKE\_BUILD\_TYPE=Sanitize. ---
**Author:** ![peter.ludemann](https://yyz2.discourse-cdn.com/free1/user_avatar/swi-prolog.discourse.group/peter.ludemann/32/48_2.png) [@peter.ludemann](https://swi-prolog.discourse.group/u/peter.ludemann)\ **Post date:** [January 3, 2026, 9:05pm UTC](https://swi-prolog.discourse.group/t/cpp2-exceptions/6040/92 "2026-01-03T21:05:41Z")
> [@mgondan1](#): > > It seems that I can, well, “fix” the problem with `LD_PRELOAD= ...` I think I’ve encountered a similar problem in the past that was solved with `LD_PRELOAD` … I forget the details but I think it was something to do with an executable requiring one path for its shared libraries but ASAN requiring a different path; and therefore the runtime ASAN library doesn’t match what the executable was built with. (So, this is something that static linking would probably solve) I’ll see if I have any notes from this and if there’s a better solution (it was years ago). Anyway, it appears that the crash is due to an ASAN shared library problem and not due to a bug in either the SWI-cpp API or some subtlety in C++ exceptions. [Previous page](https://swi-prolog.discourse.group/t/cpp2-exceptions/6040.md?page=4)