#4 NJAMD reports bogus GLIBC memory leaks

[Anonymous]

For the simple program

int
main (void)
{
bindtextdomain ("quark", "test");

return 0;
}

NJAMD reports 2 memory leaks.

NJAMD totals:

Allocation totals:      2 total, 2 leaked
Leaked User Memory:     27 bytes
Peak User Memory:       27 bytes
NJAMD Overhead at peak:     7.974 kB
Peak NJAMD Overhead:        7.974 kB
Average NJAMD Overhead:     3.986 kB per alloc
Address space used:     16.000 kB
NJAMD Overhead at exit:     7.974 kB

I discussed this with Andreas Jaeger, a GLIBC
developer, and he claims that the report is correct but
bogus, since NJAMD needs to call a cleanup function
like GLIBC's mtrace does. I quote:

"It's a communication problem between NJAMD and glibc.
glibc only frees the values it allocates if this is
explictly requested - and therefore the application has
to run the subfreeres hook. NJAMD apparently does not
do this.

If you file the bug, tell them to look at glibc's file
malloc/mtrace.c to see what's done there."

It would be nice to fix this; this and a similar "leak"
in textdomain () are the only issues NJAMD reports in
my application.

Thanks!

Neil.

[Yao Zhang]

Logged In: YES
user_id=296951

Another one which I find is running the code:

#include <stdio.h>
#include <time.h>

int main(int argc, char *argv[])
{
time_t t;

t = 0;
printf("Hello, world! %s\n", ctime(&t));

return 0;
}

reports
Allocation totals: 7 total, 6 leaked
Leaked User Memory: 1.257 kB
Peak User Memory: 1.565 kB
NJAMD Overhead at peak: 10.435 kB
Peak NJAMD Overhead: 22.743 kB
Average NJAMD Overhead: 3.770 kB per alloc
Address space used: 56.000 kB
NJAMD Overhead at exit: 22.743 kB

The cause is ctime() uses static memory but njamd should
not report memory leaks in this case.