I am trying to run the "Threading Error Analysis" on an 64bit MFC application which has a quite huge memory footprint right from the start, mostly a lot of tiny allocations for various dialogues. I'm using the 2018 Update 1 version of Intel Inspector on a Fall CU Windows 10.
When I try to start this application, "Last recorded memory usage before collection completed:" is constantly at 4095MB before the application crashes at random code locations. It appears pretty much random where it fails, but it reliably happens at the 4095MB boundary. The inspector is at this point not done loading modules yet.
The Inspector occasionally reports a "Error: Internal error. Please contact Intel customer support team." at this point.
<msg id='178' source='ccrt'> <type>INTERNAL</type> <severity>ERROR</severity> <timestamp>Wed Dec 6 09:12:30 2017</timestamp> <!-- Threadchecker or PIN generated an exception: Exception Code: RECEIVED_ACCESS_FAULT. Exception Address = 0x7ff99008b70b. Access Type: READ. Access Address = 0x7ff900000960 --> </msg> <!-- Frames of pintool call stack: --> <!-- => tclcore.dll!0x1b70b (0x7ff99008b70b) -->
Other times the application just crashes first, and Inspector doesn't notice.
The stack traces when I dump the memory are just garbage to me with Inspector attached (modules can't be resolved by Visual Studio debugger), so I can't exactly narrow this down to anything.
Is there potentially any behavior in the application itself I would need to patch out, or is this "expected" behavior from the Inspector tool?
Narrowed it down. It appears that:
Not going to bother with this crash further, I suppose it's to be expected that Internet Explorer 3 components are crashing analysis tools.