test: fix DWARF string_view linker error in blockfilter_tests #25120

1. 
   jonatack commented at 6:53 pm on May 12, 2022: member

   Fix the following build error that I am seeing with Debian testing clang 15:

   0/usr/bin/ld: /usr/bin/ld: DWARF error: invalid or unhandled FORM value: 0x23
   1test/test_bitcoin-blockfilter_tests.o: in function `blockfilter_tests::blockfilters_json_test::test_method()':
   2blockfilter_tests.cpp:(.text+0x59d6): undefined reference to `ParseHex(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)'
   3/usr/bin/ld: blockfilter_tests.cpp:(.text+0x5e54): undefined reference to `ParseHex(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)'
   4clang: error: linker command failed with exit code 1 (use -v to see invocation)
   

2. Fix DWARF string_view linker error in blockfilter_tests 7d630dc1bd
3. 
   jonatack commented at 6:57 pm on May 12, 2022: member

   Unsure, possibly related to c1d165a8c2678c31ac or to changes in clang.
4. 
   laanwj commented at 7:32 pm on May 12, 2022: member

   I’m really confused by this. Why would an explicit stringview{} be needed. Looks like the compiler thinks that ParseHex(const std::string&) is still defined where it’s not. Normally this happens if the header and compilation is out of sync.

   Did you build from a clean tree? Could this be some kind of ABI conflict between different C++ versions?

5. laanwj added the label Upstream on May 12, 2022
6. laanwj added the label Tests on May 12, 2022
7. 
   jonatack commented at 7:37 pm on May 12, 2022: member

   Yes, I suspected it was something on my end or a clang update (I updated today). Saw it several times after make distclean && ./autogen.sh && ./configure --enable-debug --with-incompatible-bdb --enable-werror --enable-c++20 --enable-suppress-external-warnings --enable-experimental-util-chainstate CC=clang CXX=clang++ && make clean && make. Anything else I should be doing to ensure a clean tree? Am building with gcc right now to compare.
8. 
   sipa commented at 7:44 pm on May 12, 2022: member

   @jonatack To be absolutely sure, you could clear the ccache cache (ccache -C) if you’re using ccache.
9. 
   laanwj commented at 7:44 pm on May 12, 2022: member

   The surest way to get a clean tree would be git clean -f -x -d (this literally wipes everything in the tree that doesn’t belong to git, so be careful you don’t have anything else in there)

   But make clean should get rid of blockfilter_tests.o. Do you use ccache? maybe also ccache -C.

10. 
    jonatack commented at 7:51 pm on May 12, 2022: member

    Thanks! The gcc build finished cleanly. Ran ccache -C and rebuilding with clang now.
11. 
    fanquake commented at 8:15 pm on May 12, 2022: member

    Yes, I suspected it was something on my end or a clang update (I updated today).

    Going to close, as this is pretty clearly a local issue. Feel free to continue discussion in regards to cleaning your git tree.

12. fanquake closed this on May 12, 2022

    ---

13. 
    jonatack commented at 8:27 pm on May 12, 2022: member

    Clearing ccache did it. Closing and sorry for the noise.
14. jonatack deleted the branch on May 12, 2022
15. 
    jonatack commented at 3:36 pm on May 13, 2022: member

    Put the lesson to use with https://github.com/jonatack/jonatack.github.io/commit/5c3169fc29d40d9a6b63dc25b4467d2ddfa1f9b7 to add a basic troubleshooting section to https://jonatack.github.io/articles/how-to-compile-bitcoin-core-and-run-the-tests.
16. fanquake locked this on May 18, 2022

Contributors
jonatack laanwj sipa fanquake

Labels
Tests Upstream