 |
|
 |
Thanks for the report Steve, giving high-priority to this |
|
|
Hi guys, so in addition to the above change, one of our tech directors is investigating other stomps with other tools and has identified the following stomp (I've attached an image). The change is the P4 diff. Unfortunately I didn't investigate this stomp so I don't have any more info on it. It looks like erasing the clock pointer inside RemoveMasterClock, and then we try to write mTarget to 0 on a now-erased clock. The fix is to zero out the target on the clock before it's erased from the mMasterClocks vector. image.png (229,895 bytes) |
 |
That change is not correct because there is another use of the RemoveMasterClock (in the method OnClockFinished) that requires later use of clock->mTarget. The easiest way to fix it, before I analyze this better is to keep a reference to the clock in RemoveUnusedMasterClocks: void TimeManager::RemoveUnusedMasterClocks(const ClocksSet& clocks)
{
for (Clock* masterClock: clocks)
{
if (!IsClockUsed(masterClock))
{
Ptr<Clock> masterClockPtr(masterClock);
RemoveMasterClock(masterClock);
masterClock->mTarget = 0;
}
}
}Could you try that? |
|
|
Hi Sergio, your suggestion works great. We've submitted that to our local branch, much appreciated. |
|
|
Both issues solved in changeset r11317. |
- Anonymous
