[Valgrind-developers] data/bss sections

[Tom H. <to...@co...>]

The debuginfo code has some curious kludge code in read_elf_debug_info 
which ignores the (correctly calculated) bss details and replaces them 
with a kludged version which makes certain (invalid) assumptions.

Does anybody know why it does this?

A related problem is that all the code for detecting global variables is 
assuming that they are in the data section and applying the data bias to 
them, which is bogus if that are really in the bss section. Of course 
the kludge currently means that the bss bias is the same anyway, but 
that is itself wrong...

Tom

-- 
Tom Hughes (to...@co...)
http://www.compton.nu/