[Lldb-commits] [lldb] [lldb] Fix nullptr dereference on running x86 binary with x86-disabled llvm (PR #82603)

[Jason Molenda via lldb-commits]

jasonmolenda wrote:

I'm a little curious the use case that led to this failure.  We have a build of llvm/lldb without the X86 target, and we're using that lldb to debug an i386/x86_64 target (gdb connection or corefile)?  Because we can't use instruction emulation based unwinding, if we don't have eh_frame for every function, backtraces will go poorly.  Once we're off of the currently-executing stack frame, we can follow the frame pointer / pc spilled to the stack (assuming the code wasn't compiled omit-frame-pointer), but we won't be able to find any other spilled registers for printing variable contents in non-volatile / caller spilled.   (and getting off of frame 0 without eh_frame is going to fail if we're in the prologue/epilogue on a thread)

https://github.com/llvm/llvm-project/pull/82603