Re: Valgrind and Bison memory leaks

[Paul Eggert]

"Joel E. Denny" <address@hidden> writes:

> I try to use Valgrind to make sure I don't create new memory leaks in 
> Bison.  However, Valgrind's complaints about all the existing memory leaks 
> really get in the way.

My kneejerk reaction is that these changes make the code harder to
follow, harder to maintain, bigger, and slower.  (I was particularly
bemused by the union of const char * and char *.  :-)

Surely there's a better way to pacify Valgrind.  Can we tell Valgrind
which leaks to ignore, in some fairly-systematic way, without needing
a lot of makework changes to Bison proper?