labath wrote: > Looks good to me, we should clear them on both attach & launch which is what > the Darwin DynamicLoader plugin was doing. > > I was a little surprised that Minidump is picking up the host native > DyanmicLoader plugin at all - seems like the static dynamic loader might make > more sense, so it can say what binaries are loaded & where they're loaded - > but I've never looked at the Minidump code, I'm sure there are reasons why it > works this way.
It originally did that, but then this changed last year in order to get access to thread-local data, which is also a responsibility of the dynamic loader class. At least for posix systems (I'm not sure about Darwin, as the dynamic loader is more tightly integrated into the OS) it also kind of makes sense because the dynamic loader can work off of an (elf) core file (if it contains enough information). For minidumps, some of that work is wasted though, because the minidump contains explicit load addresses of all modules. So, it would kind of make sense to separate the logic for loading the modules from "loading of TLS", but it's not quite clear how to do that as two are quite intertwined. https://github.com/llvm/llvm-project/pull/138892 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
