trying to understand the effect of the -init/-fini option

[Shankar Easwaran <shankarke at gmail dot com>]

I am trying to understand the effect of the -init/-fini command line 
options. I was thinking that these options create an .init section/.fini 
section and call the functions specified in the command line when the binary 
is initialized / atexit time. 

Below is an example that I am working with :- 

1.c 
---- 
int fn() { 
printf("Hello world\n"); 
return 0; 
} 

int main() { return 0; } 

$gcc --version 
gcc (Ubuntu 4.8.1-2ubuntu1~10.04.1) 4.8.1 

Case (1) 
------------ 
$gcc 1.c -Wl,-init,fn -o out 
$./out ==> Expect to print "Hello World" 
(Nothing is printed) 

Case (2) 
----------- 
$cat 1.c 
int fn() { 
return 0; 
} 

int main() { return 0; } 

$gcc -c 1.c 
$ld 1.o -init fn -fini fn -o out 
$readelf -S -W out | grep -E 'init|fini' 

(Nothing is printed) 

I expected to see the linker to create init/fini sections. 

Case (3) 
----------- 
$cat 1.c 
int fn() { 
return 0; 
} 
$gcc -c 1.c && ar cr lib1.a 1.o 
$ld -init fn lib1.a -o out -t 
(Nothing is printed) 

Expected 1.o to be pulled in from the archive and the linker to create 
init/fini sections. 

Case (4) 
----------- 
What is the difference between .init section and .init_array section ? 

When does the linker try to create either of them ? 

Similiar is the question of the difference between the .fini section and the 
.fini_array section ? 

Case (5) 
------------- 
Is there a difference when a shared object is created for the above cases as 
I see the linker is setting DT_INIT/DT_FINI to point to the functions that I 
specify for -init/-fini respectively. 


Thanks 

Shankar Easwaran