static functions appearing in the symbol table

static functions appearing in the symbol table

[Martin Faltesek]

While using gcc, I've noticed that some static functions appear in the
symbol table while others do not. What are the conditions that control
when a static function's symbol appears? I would have thought all
statics would not be included.

This observation arose while debugging stack traces in the linux
kernel. (and using ftrace).

Using arm-linux-gcc 4.9.1.

Thanks

Re: static functions appearing in the symbol table

[Martin Sebor]

On 08/23/2015 07:15 AM, Martin Faltesek wrote:
> While using gcc, I've noticed that some static functions appear in the
> symbol table while others do not. What are the conditions that control
> when a static function's symbol appears? I would have thought all
> statics would not be included.

At a high level, a static function whose calls in a program are all
inlined will not have an entry in the symbol table. A static function
that's not completely inlined will. What is and isn't inlined depends
on optimizations in effect and on attributes (e.g., function with the
always_inline attribute will likely not a have a symbol even at -O0,
while a trivial static function that's declared with attribute
noinline will even at -O1 and higher).

Martin

Re: static functions appearing in the symbol table

[Nicholas Mc Guire]

On Sun, 23 Aug 2015, Martin Faltesek wrote:

> While using gcc, I've noticed that some static functions appear in the
> symbol table while others do not. What are the conditions that control
> when a static function's symbol appears? I would have thought all
> statics would not be included.
> 
> This observation arose while debugging stack traces in the linux
> kernel. (and using ftrace).
> 
> Using arm-linux-gcc 4.9.1.
>
check your kernel config - the linux kernel has a config option 
in kernel hacking called CONFIG_OPTIMIZE_INLINING:

Kernel hacking  --->
  ...
  [*] Allow gcc to uninline functions marked 'inline' 

<snip - form the help text>
 This option determines if the kernel forces gcc to inline the functions
 developers have marked 'inline'. Doing so takes away freedom from gcc to 
 do what it thinks is best, which is desirable for the gcc 3.x series of 
 compilers. The gcc 4.x series have a rewritten inlining algorithm and 
 enabling this option will generate a smaller kernel there. Hopefully 
 this algorithm is so good that allowing gcc 4.x and above to make the 
 decision will become the default in the future. Until then this option
 is there to test gcc for this. 
<snip>

 so I suspect that this is what is causing the discrepencey between functions 
marked inline in the code and actually being compiled in. This is though
linux kernel specific and not actually a gcc issue.

also check out "3.10 Options That Control Optimization" on -finline-limit=N
and the way it is used to automatically control inlining.

thx!
hofrat