dtrace discuss - Aug 2009 - When is a string constant not a string constant?

AKA, what''s wrong with the following? Isn''t a string literal
always a constant?

$ cat > cfretain_calayer.d
pid$target::CFRetain:entry
{
    isaptr = *(uint32_t*)copyin(arg0, 4);
    classnameptr = *(uint32_t*)copyin(isaptr+8, 4);
    classname = copyinstr(classnameptr);

    p = (classname == "CALayer" ? "CFRetain(CALayer)\n" :
"");
    printf(p);
}

$ sudo dtrace -p 31303 -s cfretain_calayer.d 
dtrace: failed to compile script cfretain_calayer.d: line 8: printf( ) argument
#1 is incompatible with prototype:
	prototype: string constant
	 argument: string

Also, please let me know if there''s a better way to be doing this
(I''m a dtrace newbie).

Thanks,
Hamish
-- 
This message posted from opensolaris.org

Hey Hamish,
> AKA, what''s wrong with the following? Isn''t a string
literal always
> a constant?
>
> $ cat > cfretain_calayer.d
> pid$target::CFRetain:entry
> {
>    isaptr = *(uint32_t*)copyin(arg0, 4);
>    classnameptr = *(uint32_t*)copyin(isaptr+8, 4);
>    classname = copyinstr(classnameptr);
>
>    p = (classname == "CALayer" ? "CFRetain(CALayer)\n"
: "");
>    printf(p);
> }
>
> $ sudo dtrace -p 31303 -s cfretain_calayer.d
> dtrace: failed to compile script cfretain_calayer.d: line 8:  
> printf( ) argument #1 is incompatible with prototype:
> 	prototype: string constant
> 	 argument: string
>
> Also, please let me know if there''s a better way to be doing this
> (I''m a dtrace newbie).

DTrace requires that you provide a constant string so that it can  
perform
type checking on additional arguments (i.e. to ensure that if you  
specify
the %d format option that you have an int to print). You have two pretty
easy ways to fix this:

	printf("%s\n", classname == "CALayer" ?
"CFRetain(CALayer)" : "");

or

	trace(classname == "CALayer" ? "CFRetain(CALayer)" :
"");

Note also, that all of your variables should probably be probe-local
(prefixed with this->).

Further, if you only want trace when classname == "CALayer", rather
than
printing out an empty string, you could do something like this:

pid$target::CFRetain:entry
{
	this->isaptr = *(uint32_t*)copyin(arg0, 4);
	this->classnameptr = *(uint32_t*)copyin(isaptr+8, 4);
	this->classname = copyinstr(classnameptr);
}

pid$target::CFRetain:entry
/this->classname == "CALayer"/
{
	trace("CFRetain(CALayer)");
}

Hope that helps.

Adam

--
Adam Leventhal, Fishworks                        http://blogs.sun.com/ahl

Hamish <hamish at gmail.com> wrote:
> AKA, what''s wrong with the following? Isn''t a string
literal always a constant?
>
> $ cat > cfretain_calayer.d
> pid$target::CFRetain:entry
> {
>     isaptr = *(uint32_t*)copyin(arg0, 4);
>     classnameptr = *(uint32_t*)copyin(isaptr+8, 4);
>     classname = copyinstr(classnameptr);
>
>     p = (classname == "CALayer" ? "CFRetain(CALayer)\n"
: "");
>     printf(p);
> }
>
> $ sudo dtrace -p 31303 -s cfretain_calayer.d 
> dtrace: failed to compile script cfretain_calayer.d: line 8: printf( )
argument #1 is incompatible with prototype:
> 	prototype: string constant
> 	 argument: string
It is not simple for a compiler to check this
as p is definitely not a string constant but only pointing to one.


J?rg

-- 
 EMail:joerg at schily.isdn.cs.tu-berlin.de (home) J?rg Schilling D-13353 Berlin
       js at cs.tu-berlin.de                (uni)  
       joerg.schilling at fokus.fraunhofer.de (work) Blog:
http://schily.blogspot.com/
 URL:  http://cdrecord.berlios.de/private/ ftp://ftp.berlios.de/pub/schily

Brilliant, thank you!
-- 
This message posted from opensolaris.org

I should probably have mentioned that I only introduced the extra variable
"p" as a result of getting exactly the same error with the more direct
approach:

printf(classname == "CALayer" ? "CFRetain(CALayer)\n" :
"");

Anyway, thanks again to Adam for providing a clear solution!
-- 
This message posted from opensolaris.org