🚀 go-pugleaf

In article <1993Jan23.150923.12794@ousrvr.oulu.fi> artoinas@phoenix.oulu.fi writes:

>
>I am programming an application with Turbo C++ for Windows 3.0. I use Windows
>version 3.0. So far, I've detected all bugs and fixed them. Now one lurking
>bug is driving me mad, I have no idea about it's causes. 
>
>Well, I know that this is too general, but here is what the bug does:
>Sometimes after running the application Windows gives me an Unrecoverable
>Application error when I toggle the mouse on text editing window (in Turbo
>C++) and start typing something. However (of course) this doesn't happen al-
>ways. So if you have seen this problem and solved it, please answer?
>
>Arto Oinas - first year Computer Science student from Finland
>(Too general really I bet)
>
>

I suggest you try and run your application under Windows 3.1, if at all
possible.  That way the UAE will be replaced by a General Protection Fault,
which has the advantage of showing you *where* your program crashed.  Armed
with that address and a linker map you can then find out which function
the problem is occurring in, and track it down from there.

Best wishes,

Chris
-- 
--------------------------------------------------------------------------
| Chris Marriott                           | chris@chrism.demon.co.uk    |
| Warrington, UK                           | BIX: cmarriott              |
|                                          | CIX: cmarriott              |
| Save the whales.  Collect the whole set. | CompuServe: 100113,1140     |
--------------------------------------------------------------------------
 

artoinas@phoenix.oulu.fi (Arto Oinas) writes:


>I am programming an application with Turbo C++ for Windows 3.0. I use Windows
>version 3.0. So far, I've detected all bugs and fixed them. Now one lurking
>bug is driving me mad, I have no idea about it's causes. 

>Well, I know that this is too general, but here is what the bug does:
>Sometimes after running the application Windows gives me an Unrecoverable
>Application error when I toggle the mouse on text editing window (in Turbo
>C++) and start typing something. However (of course) this doesn't happen al-
>ways. So if you have seen this problem and solved it, please answer?

>Arto Oinas - first year Computer Science student from Finland
>(Too general really I bet)

This is a pretty general problem so try this general answer:

Check your hdc and hwnd parameters in functions like EndPaint.

I have mixed these up on numerous occasions and it has given exactly
the type of fault you describe, sometimes taking hours to sort out!
I can't understand why the STRICT compilation doesn't spot this.

Simon

In article <C1Gp35.9Ez@sunlab1.bath.ac.uk> mssshb@sunlab1.bath.ac.uk (S H Bowman) writes:
>artoinas@phoenix.oulu.fi (Arto Oinas) writes:
>
[...stuff about some bug deleted...]

>This is a pretty general problem so try this general answer:
>
>Check your hdc and hwnd parameters in functions like EndPaint.
>
>I have mixed these up on numerous occasions and it has given exactly
>the type of fault you describe, sometimes taking hours to sort out!
>I can't understand why the STRICT compilation doesn't spot this.

	This is obvious, but it's worth a check - does your
	#define STRICT appear *before* #include <windows.h>???

	Well EndPaint doesn't have an HDC, but ReleaseDC has both
	HWND & HDC.  I just stuck the following in my code somewhere:
		{HWND hwnd;
		HDC hdc;
		ReleaseDC(hdc,hwnd);	//this is WRONG
		}

	and BC++ 3.1 gives me two "suspicious pointer conversion"
	errors (also the obvious warning that hdc & hwnd are used
	before being defined).  If I switch it around:
		ReleaseDC(hwnd,hdc);	//this is RIGHT
	then I don't get those 2 pointer warnings.

	It looks like STRICT is working for me, but then again I
	have *every* single warning message turned on and I don't
	go any further until they have all been fixed.  I can't
	stress this enough for all of you programmers out there.
	My app is 22k lines of code and I don't get a single warning
	message when I do a full build (with all warnings turned on).

	I only used #pragma argsused in a couple of places where I
	have some unused function parameters in place (for future
	expansion).  Everything is fully prototyped and as bugtight
	as I could make it, and I *still* get the odd bug or two
	(as expected).

John A. Grant						jagrant@emr1.emr.ca
Airborne Geophysics
Geological Survey of Canada, Ottawa
-- 
John A. Grant						jagrant@emr1.emr.ca
Airborne Geophysics
Geological Survey of Canada, Ottawa