gdb is not calculating symbol addresses correctly for code compiled with 3.3.1

[Stephen Smith]

I am noticing this issue with binaries that are compiled with 3.3.1.  If 
the same code is compiled with 2.95 the symbols are calculated correctly.

I am debugging an embedded application and have a library that is loaded 
into memory.  I am using the add symbol file command to load its symbols 
and besides the required text section address I am specifiying the 
.data, .sdata2, .sdata, .sbss and .bss section addresses.  gdb is 
acknowledging the addresses I give it. 

Then when I ask for the address of one of the variables, I am getting an 
address totally outside the range that I specified.  If I browse the 
memory (i.e. I don't refer to the variable, but rather print the data at 
the hex address in question) I do find that the variable is at the 
address I thought it was - not the location that gdb calculated. 

Has anyone else seen this behavior?