Re: gdb-internal: determining the type of a variable

[Elena Zannoni <ezannoni at redhat dot com>]

Roul Oldenburger writes:
 > Hello everybody,
 > 
 > I hope to find advice here although this might be a gcc or 'stabs 
 > debuginformation format' related question.
 > 
 > gdb can show lots of information for a variable or instance of a 
 > structure type but I need to know where and how it gets it from.
 > 
 > Our executables are compiled with at least stabs debuginformation and I 
 > use 'objdump --stabs my_exe' to make use of it.
 > The aim is to reconstruct arbitrarily nested structure type descriptions 
 > down to their final components. I try doing this by parsing objdump's 
 > output.
 > 
 > Given the full name of a structure type inlcuding the name of the 
 > package where it is defined I will get the description of it.
 > If the components are from other structure types I can determine them 
 > quit easily through the given typenumber/-identifier, but if the 
 > components are from a traditional basic type or a self-defined basic 
 > type I am not able to determine its type clearly.
 > 
 > As an example I have the description of a structure 'sonar_description' 
 > given in the package 'awu_siso_shared_memory':
 > 
 > 73884  LSYM   0      0      00000000 1565985 
 > awu_siso_shared_memory__sonar_description:T(0,426)=s2500status:(0,123),0,8;\
 > 73885  LSYM   0      0      00000000 1566062 
 > ref_count:(0,5),32,32;id:(0,7),64,32;transmit_method:(0,139),96,8;\
 > 73886  LSYM   0      0      00000000 1566130 
 > transmit_power_level:(0,397),128,32;measuring_range:(0,382),160,384;\
 > 73887  LSYM   0      0      00000000 1566200 
 > search_sector:(0,400),544,32;search_rotation:(0,148),576,8;\
 > 73888  LSYM   0      0      00000000 1566261 
 > cw_pulse:(0,425),608,32;multi_number_of:(0,7),640,32;\
 > 73889  LSYM   0      0      00000000 1566316 
 > multi_duration_of:(0,33),672,64;;
 > 
 > The given information decribes the structure, but if I want to determine 
 > of what type the component 'multi_duration_of' is (grep 
 > ':t(0,33)=[r|@]'), I will get multiple possibilities - here are some:
 > 
 > 210    LSYM   0      71     00000000 8722 
 > awu_common__debug_mode_range___XDLU_1__1:t(0,33)=@s8;r(0,27);1;1;
 > 371    LSYM   0      84     00000000 12939 
 > awu_common__types__nlink_t:t(0,33)=r(0,29);0;-1;
 > 441    LSYM   0      44     00000000 15165 
 > awu_common__stdio__TTstringS2bP1___XDLU_1__71:t(0,33)=r(0,1);1;71;
 > 6592   LSYM   0      120    00000000 184371 
 > system__address:t(0,33)=r(0,33);0;-1;
 > 19690  LSYM   0      109    00000000 425771 
 > duration___XF_1_1000000000:t(0,33)=@s64;r(0,33);01000000000000000000000;0777777777777777777777;
 > 180473 LSYM   0      120    00000000 184371
 > 122185 LSYM   0      100    00000000 2948958 
 > system__secondary_stack__Tmark_idB:t(0,33)=r(0,33);0;-1;
 > 6835   LSYM   0      81     00000000 75724 
 > awu_common__ipc__Tipc_permissionB:t(0,33)=r(0,33);-2147483648;2147483647;
 > 
 > 
 > In fact the type is duration, but my problem is to distinguish between 
 > the different possibilities ... so looking for the typenumber alone is 
 > not enough.
 > 
 > How does gdb does it??

I see that Ian answered your questions on the binutils list.  As far
as how gdb does it, look at stabsread.c, there is a big switch
statement that handles the cases.

elena

 > 
 > What am I doing wrong?
 > 
 > Probably looking at stabs does not give the right information to do 
 > this; or do i misinterpret stabs format?
 > 
 > Thanks for any comments and help!
 > 
 > Roul Oldenburger