Re: pmmgr mem leak or someone else's problem?

[fche@xxxxxxxxxx (Frank Ch. Eigler)]

kenj wrote:

> [...]
>
> === valgrind stdout === 
> ==6948== 132,000 bytes in 165 blocks are definitely lost in loss record 16 of 
> 16
> ==6948==    at 0x4C28192: operator new[](unsigned long) 
> (vg_replace_malloc.c:363)
> ==6948==    by 0x10FDC0: pmmgr_job_spec::compute_hostid(std::string const&) 
> (pmmgr.cxx:249)
> [...]

It's legitimate bug/leak.  Fix:


diff --git a/src/pmmgr/pmmgr.cxx b/src/pmmgr/pmmgr.cxx
index 26ef9fb7f248..31d2040a8a1a 100644
--- a/src/pmmgr/pmmgr.cxx
+++ b/src/pmmgr/pmmgr.cxx
@@ -278,6 +278,7 @@ pmmgr_job_spec::compute_hostid (const pcp_context_spec& ctx)
       hostid_fields.push_back (values[i].cp);
 
   (void) pmDestroyFetchGroup (fg);
+  delete[] values;
 
   // Sanitize the host-id metric values into a single string that is
   // suitable for posix-portable-filenames, and not too ugly for


- FChE