Open GoogleCodeExporter opened 9 years ago
Forgot to mention that valgrind detects the last leak, so it's probably not
problem with compilation.
Original comment by patrik...@gmail.com
on 26 Nov 2013 at 11:08
I think it's the same issue with issue
537(https://code.google.com/p/gperftools/issues/detail?id=537)
To avoid this, I use
(1)export HEAPCHECK="draconian"
(2)run program
(3)pprof ./test.bin "/tmp/test.bin.57880._main_-end.heap"
--ignore='_dl_rtld_di_serinfo' --lines --show_bytes --pdf > 1.pdf
As draconian will report gperftools objects as leak too.
Original comment by xiaoyur...@gmail.com
on 1 Jan 2014 at 11:52
Interesting, on my FreeBSD system the original testcase shows "detected leaks
of 120 bytes in 3 objects"
But when I place "printf("something\n")" after each malloc, output changes:
"detected leaks of 280 bytes in 7 objects"
Something isn't quite right.
Original comment by yuriv...@gmail.com
on 28 Aug 2014 at 9:40
Additionally, result also depends on compiler. My previous result was with
clang.
With gcc even with printf it reports "detected leaks of 160 bytes in 4 objects
" (vs. correct result with clang)
Original comment by yuriv...@gmail.com
on 28 Aug 2014 at 9:52
Original issue reported on code.google.com by
patrik...@gmail.com
on 26 Nov 2013 at 10:04Attachments: