[RESOLVED] DGindexNV.exe starting poorly

[Sharc]

In the cmd window the cursor just returns. No error message, nothing.

[admin]

Yes, just like AVG. Can you program exceptions with Windows Defender?

Could they be detecting my detecting them and deciding that is suspicious behavior? Where is my tin-foil hat when I need it?



BTW, DGAVCDec does not contain any such anti-reversing code.

[Sharc]

Nice helmets! Reminds me of Saul's partner .....

And yes, putting DGindexNV.exe in Defender's exception list eliminated the problem. Thanks.

Edit:
Too early joy .... the problem is back. Even disabling Defender does not help. Must be something else.
Rest assured that I have no reverse-engineering tool installed - at least not as I know.

Here a sequence of starting attempts (g=good, b=bad):
g b b b g g g g g g g b g g g g g b g b g g g b g g g g g g b g b b g g g b g ..........

[admin]

"I can't reproduce it, it seems random."

I hate that random stuff. You have to start an exhaustive search to find it or get lucky. Maybe we can get lucky.

I have a first debug build for you.

First, get the DbgView utility from Microsoft:

https://technet.microsoft.com/en-us/sys ... gview.aspx

Open it. Pull down Edit/Filter Highlight. Change the Include field to DGIndexNV instead of *. Then OK. Then Edit/Clear Display and leave it open.

Then get the test build (32 bit):

http://rationalqm.us/misc/DGIndexNV_Sharc.exe

Then make the problem happen and tell me if any messages are generated in the Dbgview window.

Thank you.

[Sharc]

When DGIndexNV_Sharc.exe starts correctly I get in the debug window like:
33 213.08460999 [9236] DGIndexNV ref 5

When it fails to start I get nothing.

[admin]

You shouldn't see ref 5 until you exit. Is that the case?

Standby for another test build...

[admin]

Please re-download the test build and report the results.

[Sharc]

When it starts OR fails to start:
ref 10
ref 11
ref 12
ref 13

ref 5 when it exits after a successful start

[admin]

OK, that shows that the OS is at least starting DGIndexNV. Now we have to see where and why it is killing itself.

Please re-download and test again. Thank you.

[Sharc]

Successful start:
ref 10 ..... ref 21
Exit:
ref 5

Fail to start:
ref 10 ..... ref 15 (ref 16 ... ref 21 are missing)

[Sharc]

No rush. Enjoy the pool. (It's already dark and cold here....)

[admin]

Getting closer. Please redownload and test again.

I'll have to be away coaching at the pool for a few hours starting at about 12:20. But I'll get right back on it at my return. We should be able to nail this down with a few more iterations.

[Sharc]

Successful:
ref 10,11,12,13,14,15,16,17,18,19,20,21,40,41,42 Exit: ref 5

Failure:
ref 10,11,12,13,14,15

[admin]

Oh, ExitProcess(), that's how I punish reverse engineers, but here it's the honorable Sharc, so I must have messed up.

Aha, I think I've got it! Standby for a test build with a fix.

Uninitialised local variable. I must have been ethanol-coding that day and drifted slightly below the sweet spot. A slightly higher BAC would have avoided this failure. https://www.targetprocess.com/blog/2014 ... love-beer/

[admin]

OK, please redownload the Sharc test build and report. I am optimistic that it is fixed.

If it's OK I will make a slipstream.

[Sharc]

Looks good, yes:

Start: ref 10,11,12,13,14,15,16,17,18,19,20,21 Exit: ref 40,41,42,5

[admin]

Thank you, Sharc. Just to make sure you are not beer-testing outside the sweet spot, I will wait a few hours and then make a slipstream. Thank you so much for your debugging and reporting skills.

[Sharc]

My pleasure. You are most welcome

[admin]

"The peak of the curve occurs at a BAC of 0.1337%, which is a reference to leet."

http://www.explainxkcd.com/wiki/index.p ... llmer_Peak

I agree that the real peak is much less than 0.1337, which would be considered highly impaired for driving.

[CarlEdman]

Thanks to Sharc and DG for fixing this bug!

I've been encountering it a bunch of times of late and this morning I finally isolated a a reproducible set of test conditions and was just about to make a new thread with a detailed description and copies of the various files involved. Fortunately, I first checked the new binaries thread, noticed the new version from exactly 2 weeks ago, tried it and--wouldn't you know it--it squashed exactly the bug I was about to report.

[admin]

Thanks for your additional test results, Carl.