dtrace discuss - Aug 2006 - Why does dtrace -h remove const qualifier?

If I have the following definition:
provider foo {
   probe bar(const char*);
};

and creates the headerfile (snv 43)  with dtrace -h I get:
extern void __dtrace_foo___bar(char *);

This is a "problem" for me (ok, I could add an explicit cast each time
I fire the probe, or edit the generated headerfile by hand...), since it
generates compilation errors when I call the probe in a C++ context:
"a.cc", line 6: Error: Formal argument 1 of type char* in call to
__dtrace_foo___bar(char*) is being passed const char*.

So, is this a bug in dtrace or is there a valid reason for removing the
constness in the function prototype...

Trond
 
 
This message posted from opensolaris.org

Hey Trond,

This is a bug in libdtrace. The problem is that the DTrace compiler drops
const declarations. The two solutions we''ve thought of for this problem
are to enhance the compiler (a bunch of work) or make the prototypes in
the generated header files _all_ be const (simpler, but not as simple as
it had originally seemed).

Go ahead and file the bug in library/libdtrace and I''ll add some notes
to
the bug report. I''ve been meaning to put together a page of ways people
from the community could improve DTrace -- this bug would be on that list.

Adam

On Mon, Aug 14, 2006 at 12:00:35AM -0700, Trond Norbye
wrote:
> If I have the following definition:
> provider foo {
>    probe bar(const char*);
> };
> 
> and creates the headerfile (snv 43)  with dtrace -h I get:
> extern void __dtrace_foo___bar(char *);
> 
> This is a "problem" for me (ok, I could add an explicit cast each
time I fire the probe, or edit the generated headerfile by hand...), since it
generates compilation errors when I call the probe in a C++ context:
> "a.cc", line 6: Error: Formal argument 1 of type char* in call to
__dtrace_foo___bar(char*) is being passed const char*.
> 
> So, is this a bug in dtrace or is there a valid reason for removing the
constness in the function prototype...
> 
> Trond
>  
>  
> This message posted from opensolaris.org
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
-- 
Adam Leventhal, Solaris Kernel Development       http://blogs.sun.com/ahl