I've been taking a look at the memory leak in t-locale. According to
the valgrind log there is a leak due to operator new(unsigned long) in
set_point when called from check_output().
Here is the relevant code:
char point_string[2];
#if HAVE_STD__LOCALE
// Like std::numpunct, but with decimal_point coming from point_string
[].
class my_numpunct : public numpunct<char> {
public:
explicit my_numpunct (size_t r = 0) : numpunct<char>(r) { }
protected:
char do_decimal_point() const { return point_string[0]; }
};
#endif
void
set_point (char c)
{
point_string[0] = c;
#if HAVE_STD__LOCALE
locale loc (locale::classic(), new my_numpunct ());
locale::global (loc);
#endif
}
void
check_output (void)
{
static char point[] = {
'.', ',', 'x', '\xFF'
};
for (size_t i = 0; i < numberof (point); i++)
{
set_point (point[i]);
ostringstream got;
// <snip>
}
}
It seems to me that the "new my_numpunct ()" is the culprit. The
constructor just calls the constructor for the standard numpunct
class, whose destructor is:
explicit numpunct(
size_t _Refs = 0
);
As size_t is probably an int of 4 bytes, it looks like that never gets
cleaned up. Apparently the destructor for that class is protected.
My knowledge of C++ doesn't enable me to see a solution to this. Does
anyone else know how to force cleanup of excrement like this?
Bill.
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"mpir-devel" group.
To post to this group, send email to mpir-devel@googlegroups.com
To unsubscribe from this group, send email to
mpir-devel+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/mpir-devel?hl=en
-~----------~----~----~----~------~----~------~--~---
