18:40:48 <m-relay> <0​xfffc:monero.social> MSYS2 CI issue is upstream: https://github.com/boostorg/boost/issues/1079
19:09:35 <selsta> 0xfffc: did you see the same behavior with boost 1.88?
19:23:53 <m-relay> <0​xfffc:monero.social> selsta: Yes. I double confirmed again right now.  Another data point: look at the ICU libraries found by boost ( when building boost via pacman ), you will see it founds only `.dll.a` files.
19:24:08 <m-relay> <0​xfffc:monero.social> > FIND_PACKAGE_MESSAGE_DETAILS_ICU:INTERNAL=[C:/msys64/mingw64/include][C:/msys64/mingw64/lib/libicudt.dll.a;C:/msys64/mingw64/lib/libicuin.dll.a;C:/msys64/mingw64/lib/libicuuc.dll.a;C:/msys64/mingw64/lib/libicudt.dll.a;C:/msys64/mingw64/lib/libicuin.dll.a;C:/msys64/mingw64/lib/libicuuc.dll.a][found components: data i18n uc ][v77.1()]
19:24:18 <m-relay> <0​xfffc:monero.social> from CMakeCache.txt in boost build directory.
19:25:01 <selsta> what I meant is if it finds the same libs with 1.88 I'm not sure if this is the issue we are having as the issue for us got introduced once mingw updated boost to 1.89
19:27:03 <m-relay> <0​xfffc:monero.social> selsta: You are correct. But this is the root issue. When pacman updated boost to 1.89, it broke something.
19:27:31 <m-relay> <0​xfffc:monero.social> But the root of the issue is this one. This is the build command: https://paste.debian.net/1394883/
19:27:53 <m-relay> <0​xfffc:monero.social> when you add `.dll` to the end of icu files, it finds them and works correctly.
19:28:22 <m-relay> <0​xfffc:monero.social> `-licuio.dll -licuin.dll -licuuc.dll -licudt.dll -licutu.dll` works without issue and compiles.
19:29:09 <m-relay> <0​xfffc:monero.social> the fact that boost is in the middle of transition from b2 to CMake, and the pacman used cmake prematurely for building for msys2/mingw is make it convoluted.
19:32:30 <m-relay> <0​xfffc:monero.social> In case anyone interested: a nice and easy PR would change ICU_LIBRARIES to find_package(ICU ...) and ICU::i18n style CMake ( Iconv::Iconv too ). But that is not going to solve this issue. Since the root of the issue is the way boost handles compilation.
19:46:52 <selsta> i didn't realize they switched to cmake
19:48:22 <m-relay> <0​xfffc:monero.social> Not official / final yet. https://github.com/boostorg/cmake/ they have been actively working on this since 2020 IIRC. The  msys2/mingw should've not used it. since it is pretty early stage.  https://github.com/msys2/MINGW-packages/blob/852bf752f48933b4d38d371db3a6d519299ed7c1/mingw-w64-boost/PKGBUILD#L80
19:57:49 <selsta> maybe in addition a msys2 issue should be opened? they can quickly put out a new version
19:58:08 <selsta> though since our issue is quite specific not sure if they would care
20:11:37 <m-relay> <0​xfffc:monero.social> not a bad idea. worth the try.