|
From: Bruce L. <low...@si...> - 2007-01-12 21:29:04
|
This exists at least from 3.2.1 and is in trunk as well as branch32.
When using memcheck to check for leaks with gen-suppressions=yes and
--show-reachable=no, reachable blocks are pushed through the
record_leak_error code and unique_error code with the
print_record/print_error flag=false. The result is that no
information is printed about the reachable block, but the
==26027== ---- Print suppression ? --- [Return/N/n/Y/y/C/c] ---- y
prompt is still printed. This can obviously happen hundreds or
thousands of times in between actual unreachable blocks that might
want to be repressed in the future.
I did this to fix it:
Index: m_errormgr.c
===================================================================
--- m_errormgr.c (revision 6510)
+++ m_errormgr.c (working copy)
@@ -672,8 +672,8 @@
VG_(message)(Vg_UserMsg, "");
pp_Error(&err);
is_first_shown_context = False;
+ do_actions_on_error(&err, allow_db_attach);
}
- do_actions_on_error(&err, allow_db_attach);
return False;
and I can't imagine why one would want to attach to an unprinted error
anywhere, either, but I guess the other option would be not to push
unprinted leaks through this code.
Bruce
|
|
From: Nicholas N. <nj...@cs...> - 2007-01-13 00:03:44
|
On Fri, 12 Jan 2007, Bruce Lowekamp wrote: > This exists at least from 3.2.1 and is in trunk as well as branch32. > > When using memcheck to check for leaks with gen-suppressions=yes and > --show-reachable=no, reachable blocks are pushed through the > record_leak_error code and unique_error code with the > print_record/print_error flag=false. The result is that no > information is printed about the reachable block, but the > ==26027== ---- Print suppression ? --- [Return/N/n/Y/y/C/c] ---- y > prompt is still printed. This can obviously happen hundreds or > thousands of times in between actual unreachable blocks that might > want to be repressed in the future. > > I did this to fix it: > Index: m_errormgr.c > =================================================================== > --- m_errormgr.c (revision 6510) > +++ m_errormgr.c (working copy) > @@ -672,8 +672,8 @@ > VG_(message)(Vg_UserMsg, ""); > pp_Error(&err); > is_first_shown_context = False; > + do_actions_on_error(&err, allow_db_attach); > } > - do_actions_on_error(&err, allow_db_attach); > > return False; > > and I can't imagine why one would want to attach to an unprinted error > anywhere, either, but I guess the other option would be not to push > unprinted leaks through this code. I applied the patch, it will be in 3.2.2. Thanks. Nick |