On 2016/01/28 7:06, Philippe Waroquiers wrote:
> On Thu, 2016-01-28 at 02:19 +0900, ISHIKAWA,chiaki wrote:
>> c++11 runtime seems to use |new| operation to create some data in a very
>> primitive internal string handling function, and these string data are
>> "free"ed by many other functions that my application (mozilla
>> thunderbird) use. So delete vs free issues are reported. I suspect c++11
>> ought to use "malloc()" for the internal string operation, but then
>> maybe other parts of c++11 library may complain about malloc vs delete
>> mismatch then :-(
> Some such false positive of 'mismatched malloc/free/new/delete'
> were found in the past due to one operation being inlined
> (e.g. the new) but the delete not being inlined.
> The matching code does not understand inlining, and so the false
> positive.
>
> Maybe that is your case ?
> (you might check this by showing the stacktrace for the allocation
> and deallocation using e.g. gdb+vgdb, and see if inlining enters in
> the game).
>
> Philippe

Thank you for the tips.

I will investigate this issue further along your suggestions.
I can whitelist typical noises by using suppression file, but
to think the valgrind/memcheck needs to check this and notice the 
anomaly, check the
backtrace of calls to malloc/new to report it and only then learns it is 
whitelisted, and moves on to the original execution: such a waste of 
runtime CPU (!).
Actually, I think I already notice a significant slowdown due to this 
added mismatches.
Or maybe I am hallucinating and in need of a faster CPU :-)

But if the issue is the unbalanced inlining, there is a hope that I can 
either fix the
local build environment to inline both new/delete, etc., or try 
contacting the library builders to fix their
toolchain setup.

CI


------------------------------------------------------------------------------
Site24x7 APM Insight: Get Deep Visibility into Application Performance
APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
Monitor end-to-end web transactions and take corrective actions now
Troubleshoot faster and improve end-user experience. Signup Now!
http://pubads.g.doubleclick.net/gampad/clk?id=267308311&iu=/4140
_______________________________________________
Valgrind-users mailing list
Valgrind-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/valgrind-users

Reply via email to