Experimental Physics and
Industrial Control System

"J. Lewis Muir" <[email protected]> · Mon, 06 Apr 2015 12:37:31 -0500

On 4/6/15 10:22 AM, Andrew Johnson wrote:
> Anyone else want to comment on this idea?

I'm not experienced with Valgrind, and I'm not saying including
valgrind.h in Base is a bad idea, but my initial reaction is that it
seems lame to have to add macros and whatnot to the code to support a
tool like Valgrind.  I'd like the source code to be as clean and clear
as possible, and, to me, adding Valgrind-specific stuff definitely
doesn't help with that.

I'm wondering, why does the code need to be modified to use Valgrind?
Isn't Valgrind supposed to work without modifying the code?  The "About"
page on the Valgrind website says,

  "Valgrind is easy to use.  Valgrind uses dynamic binary
   instrumentation, so you don't need to modify, recompile or relink
   your applications.  Just prefix your command line with valgrind and
   everything works." [1]

Is it because the code uses memory pools and hence is not allocating and
deallocating memory "normally?"  If so, what about instead modifying the
code to support a way to easily disable the use of memory pools?  Then
you could run the program under Valgrind with memory pools disabled, and
there wouldn't be a need for any Valgrind-specific modifications to the
code?

Thanks,

Lewis

[1] http://valgrind.org/info/about.html

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System