Could someone more familiar with dwarf2out.c take a look at PR 44645? http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44645
It is a fairly problematic regression in C++ debug info, which causes errors in the python pretty printers for libstdc++ associative contaienrs (map, set etc.) and has several duplicate PRs now. As Tom Tromey originally identified, the DIE for some pointer types is missing DW_AT_type, which makes GDB think it's dealing with a void* not the actual pointer type, and that prevents dereferencing pointers while debugging. I've tried to find a problem in dwarf2out.c on the 4.5 branch and compared it with the versions on the 4.4 branch and trunk (which don't have the bug) but I'm stumped. Any pointers or help would be much appreciated. Jonathan
