labath wrote: > Another aside: Does anyone know why the [libstdc++ `std::__cxx11::string` > summaries](https://github.com/llvm/llvm-project/blob/802fa92aee3565768887615108aa3e924d4e0fc7/lldb/source/Plugins/Language/CPlusPlus/LibStdcpp.cpp#L240-L358) > weren't using the string summary? They were initially added in > [D13964](https://reviews.llvm.org/D13964), but it doesn't provide a reason to > do so. Note that the summary for a `std::string v = "foo"` was the format > string one (`${var._M_dataplus._M_p}`). Only when printing a reference or > pointer to a string, `LibStdcppStringSummaryProvider` was called. I tried to > use this function for all libstdc++ strings, but this broke when using > `-D_GLIBCXX_USE_CXX11_ABI=0` (on MinGW). On the other hand, > `${var._M_dataplus._M_p}` works with both ABIs.
Well, the commit says (I haven't verified this) that the std::string formatter at the time was written in python, so this could have been a way to avoid the python dependency. As for the implementation, a possible reason (again, I don't know if that's the case) could be to make the formatter work in the case where you don't have debug info for std::string -- it looks like it avoids looking at any type information, which means it could work even with a forward declaration -- but that it would break when the ABI changes, as you've noticed. I don't think we need to concern ourselves with that though. Your new implementation is fine. https://github.com/llvm/llvm-project/pull/143177 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
