Re: [hlcoders] Bizzare/Random/Frightening crash

2005-02-22 Thread NuclearFriend
I too have this problem and I too am running MSVC .Net 2002 (7.0).
Sometimes it appears, sometimes it doesn't. Sometimes I've changed
something (although nothing to do with those destructors), sometimes
I've changed nothing. I am considering upgrading to MSVC .Net 2003 but
I will first have a different coder on my mod team try and compile
this current build with their copies of 2003. I'll let you know if I
get any more info.

Jonathan Murphy
Programmer for Resistance and Liberation Mod


On Tue, 22 Feb 2005 00:39:25 -0500, Maurino Berry [EMAIL PROTECTED] wrote:
 unfortunately cleaning it doesnt seem to help very often : I think I'm
 going to try compiling on another computer maybe if I can't get any more
 suggestions from here.. thanks though :D

 _
 Designer Mail isn't just fun to send, it's fun to receive. Use special
 stationery, fonts and colors.
 http://join.msn.com/?pgmarket=en-capage=byoa/premxAPID=1994DI=1034SU=http://hotmail.com/encaHL=Market_MSNIS_Taglines
   Start enjoying all the benefits of MSN® Premium right now and get the
 first two months FREE*.

 ___
 To unsubscribe, edit your list preferences, or view the list archives, please 
 visit:
 http://list.valvesoftware.com/mailman/listinfo/hlcoders



___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlcoders



RE: [hlcoders] Bizzare/Random/Frightening crash

2005-02-22 Thread Karl \XP-Cagey\ Patrick
Inconsistent exit crashes can be caused when there are dependencies on
automatically destructed singletons or static variables--the stack you gave
looks like final cleanup, and if the call to RemoveAllPanels uses an object
that has already been destructed (possibly the CBaseViewport itself,
possibly a singleton used by that function call), the result would be
gibberish; calls into that object would have erratic behavior.

The compiler should in theory use a consistent cleanup ordering between
compiles, but there's no requirement specified by the C++ standard that
would enforce this, and if the destruction order isn't consistent for some
reason, it would explain the behavior that you're seeing from compile to
compile.  Any change to apparently unrelated code could potentially result
in the destruction order of singletons being changed.

I'd begin looking for early destruction of resources used by RemoveAllPanels
(including the CBaseViewport itself) as a cause.

Karl

-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Maurino Berry
Sent: Monday, February 21, 2005 9:39 PM
To: hlcoders@list.valvesoftware.com
Subject: Re: [hlcoders] Bizzare/Random/Frightening crash

unfortunately cleaning it doesnt seem to help very often : I think I'm
going to try compiling on another computer maybe if I can't get any more
suggestions from here.. thanks though :D

_
Designer Mail isn't just fun to send, it's fun to receive. Use special
stationery, fonts and colors.
http://join.msn.com/?pgmarket=en-capage=byoa/premxAPID=1994DI=1034SU=htt
p://hotmail.com/encaHL=Market_MSNIS_Taglines
  Start enjoying all the benefits of MSNR Premium right now and get the
first two months FREE*.


___
To unsubscribe, edit your list preferences, or view the list archives,
please visit:
http://list.valvesoftware.com/mailman/listinfo/hlcoders




___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlcoders



Re: [hlcoders] Bizzare/Random/Frightening crash

2005-02-22 Thread tei

Maurino Berry wrote:
Please bare with me while I try to explain this crash, let me begin by
saying I'm using the very first mp sdk source with no updates... and debug
build is not helping at all. Perhaps the most frightening aspect of this
crash is its consistancy between game launches, and its inconsistancy
between builds, its almost as if my compiler is messed up. Basically I'm
encountering a crash in the client dll. When I create a server, disconnect,
and then recreate. the debug build gives me an error in
[...]
-
Short version email:
Try this:
http://www.lost.co.nz/main/software/misc/mfcclean.html
-
Long version email:
C++ Annotated Comments:
 Often its easy to have any version system (CVS, etc..) to move backward
and test bugs.
 Its posible that a bug become that dificult to solve, that the better
method its to reject ANY change from the latest backup:
 v1
 v2
 v3 VERFATABUG!
 (you backtrack to v2)
 v2b
 v3
 v4
 To make v2b you can start v2 and add v3 changes on a incremental way.
This will help you hunt the dificult bug. But you may ignore the later
code, and reformulate it. Rewrite from crach that code.
 Its posible that a IDE compiler corrupt temporal intermediate files,
even makefile-driven compilers often corrupt something. You need to
determine what its your source and what its crap, to remove the crap.
Here its a NICE tool to remove the Visual Studio crap
http://www.lost.co.nz/main/software/misc/mfcclean.html
 Anyways you NEED to know how to clean your code by hand. If you dont
know what files are source code, and what files intermediate crap, you
become a lame dumbass isolated on graphic-interface-land.
Proffs!
and
Proffs!!!
 Whats its science? the Truth? naaa.. that its religion. Science its
about to proff something its broken. Something you can't proff its not
broken its not science. And science work!, this why you can use science
to build rockets to moon, but you can't use religion, sex or other stuff.
Only science.
Whats this related to code?
 A good building system its automatically, and include automatically
testing.
 If you are moding a engine, you use demo files to test bugs on netcode.
 If you are moding a client, you create a demo mod to test anything on
client and stress test.
 If you are moding a mod, you create a script/macro/something/zoomap to
test anithing the mod include.
So: make a zoo map and try to crash your mod.
Corolary:
If you can't proff something its broken. Avoit that stuff, ca be broken,
Its thick ice.
___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlcoders


RE: [hlcoders] Bizzare/Random/Frightening crash

2005-02-22 Thread Yahn Bernier
Yes, this seems to be a problem where the call stack shows a destructor
getting called on .dll unload of the client .dll.

client.dll!$E11()  + 0x13   C++

Anything that looks like $E11() is an object sitting on the stack during
.dll init.

At the lower level, it appears to be trying to delete a vgui::Panel
object of some kind, but vgui is most likely already shut down, so it's
crashing (the panel was probably already deleted by vgui during
shutdown).  You might try marking the m_pViewport (SDKViewport) panel as
SetAutoDelete( false ); and see if that resolves the crash.  Or it could
be a child panel of the viewport (a hud element of some kind).

You could probably put a breakpoint in the removeallpanels call and see
which panel it's puking on (you might have to print them out during
startup to get meaningful panel names from them...)

Yahn

-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Karl
XP-Cagey Patrick
Sent: Tuesday, February 22, 2005 6:25 AM
To: hlcoders@list.valvesoftware.com
Subject: RE: [hlcoders] Bizzare/Random/Frightening crash

Inconsistent exit crashes can be caused when there are dependencies on
automatically destructed singletons or static variables--the stack you
gave looks like final cleanup, and if the call to RemoveAllPanels uses
an object that has already been destructed (possibly the CBaseViewport
itself, possibly a singleton used by that function call), the result
would be gibberish; calls into that object would have erratic behavior.

The compiler should in theory use a consistent cleanup ordering between
compiles, but there's no requirement specified by the C++ standard that
would enforce this, and if the destruction order isn't consistent for
some reason, it would explain the behavior that you're seeing from
compile to compile.  Any change to apparently unrelated code could
potentially result in the destruction order of singletons being changed.

I'd begin looking for early destruction of resources used by
RemoveAllPanels (including the CBaseViewport itself) as a cause.

Karl

-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Maurino
Berry
Sent: Monday, February 21, 2005 9:39 PM
To: hlcoders@list.valvesoftware.com
Subject: Re: [hlcoders] Bizzare/Random/Frightening crash

unfortunately cleaning it doesnt seem to help very often : I think I'm
going to try compiling on another computer maybe if I can't get any more
suggestions from here.. thanks though :D

_
Designer Mail isn't just fun to send, it's fun to receive. Use special
stationery, fonts and colors.
http://join.msn.com/?pgmarket=en-capage=byoa/premxAPID=1994DI=1034SU
=htt
p://hotmail.com/encaHL=Market_MSNIS_Taglines
  Start enjoying all the benefits of MSNR Premium right now and get the
first two months FREE*.


___
To unsubscribe, edit your list preferences, or view the list archives,
please visit:
http://list.valvesoftware.com/mailman/listinfo/hlcoders




___
To unsubscribe, edit your list preferences, or view the list archives,
please visit:
http://list.valvesoftware.com/mailman/listinfo/hlcoders


___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlcoders



Re: [hlcoders] Bizzare/Random/Frightening crash

2005-02-21 Thread Hasan Aljudy
hmm ..
not that I know much but, sometimes builds get messed up for some
reasons (I don't think it's a bug in the compiler itself .. but what
do I know) but you can fix this kind of stuff by rebuilding a clean
build.

On Mon, 21 Feb 2005 23:38:10 -0500, Maurino Berry [EMAIL PROTECTED] wrote:
 Please bare with me while I try to explain this crash, let me begin by
 saying I'm using the very first mp sdk source with no updates... and debug
 build is not helping at all. Perhaps the most frightening aspect of this
 crash is its consistancy between game launches, and its inconsistancy
 between builds, its almost as if my compiler is messed up. Basically I'm
 encountering a crash in the client dll. When I create a server, disconnect,
 and then recreate. the debug build gives me an error in

 void CBaseViewport::RemoveAllPanels(void) (vgui_teamfortressviewport.cpp
 line 413)

 this is the callstack..

client.dll!CBaseViewport::RemoveAllPanels()  Line 413 + 0xf C++
client.dll!CBaseViewport::~CBaseViewport()  Line 432C++
client.dll!SDKViewport::~SDKViewport()  + 0x14  C++
client.dll!SDKViewport::`scalar deleting destructor'()  + 0x14  C++
client.dll!ClientModeShared::~ClientModeShared()  Line 107 + 0x22  
  C++
client.dll!ClientModeSDKNormal::~ClientModeSDKNormal()  Line 92 + 0x8  
  C++
client.dll!$E11()  + 0x13   C++
client.dll!doexit(int code=0, int quick=0, int retcaller=1)  Line 376  
  C
client.dll!_cexit()  Line 318 + 0xb C
client.dll!_CRT_INIT(void * hDllHandle=0x2400, unsigned long
 dwReason=0, void * lpreserved=0x0001)  Line 193 C
client.dll!_DllMainCRTStartup(void * hDllHandle=0x2400, unsigned 
 long
 dwReason=0, void * lpreserved=0x0001)  Line 312 + 0x11  C

 The kicker is the call stack indicates the call to this method is happening
 after a call to doExit() which is totally unrelated to anything i've been
 doing - if thats not enough for you, it gets a whole lot worse. I ported one
 single npc over nothing more, and magically the crash was fixed. Now I can
 create/shutdown servers as many times as I want without crashing

 It gets worse.

 Now when creating an npc I experience the exact same crash. then maybe 30
 rebuilds later.. the crash is magically gone, without even going near npc or
 combat character code of anykind... another few builds later, the crash
 reappears. (totally random)

 Keep in mind the crashing is _consistant_ in a build, as in once the crash
 is present no matter how many times i launch the game I'll have the exact
 same behaviour, magically sometimes if I do a rebuild without changing any
 code, the crash vanishes. I checked this to make sure it wasn't something
 like accessing an invalid pointer.

 I wasn't even going to post about it here but I thought I'd give it a shot.
 This must be something compiler related (mvsc++7.0) OR for some messed up
 reason I couldn't understand the debugger is showing me the same (old and
 unrelated) callstack, however I don't see how this could be the case as the
 debugger shows the proper information if I crash a different way.

 Does *anyone* have *any* suggestions? am I missing some service pack I don't
 know about?
 Thanks in advance...

 _
 Powerful Parental Controls Let your child discover the best the Internet has
 to offer.
 http://join.msn.com/?pgmarket=en-capage=byoa/premxAPID=1994DI=1034SU=http://hotmail.com/encaHL=Market_MSNIS_Taglines
  Start enjoying all the benefits of MSN® Premium right now and get the
 first two months FREE*.

 ___
 To unsubscribe, edit your list preferences, or view the list archives, please 
 visit:
 http://list.valvesoftware.com/mailman/listinfo/hlcoders



___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlcoders



Re: [hlcoders] Bizzare/Random/Frightening crash

2005-02-21 Thread Luke Graham
This really sounds like a build system problem. The rebuild without
changing code fix is what gives it away. Make clean and make.


On Mon, 21 Feb 2005 21:55:23 -0700, Hasan Aljudy [EMAIL PROTECTED] wrote:
 hmm ..
 not that I know much but, sometimes builds get messed up for some
 reasons (I don't think it's a bug in the compiler itself .. but what
 do I know) but you can fix this kind of stuff by rebuilding a clean
 build.

 On Mon, 21 Feb 2005 23:38:10 -0500, Maurino Berry [EMAIL PROTECTED] wrote:
  Please bare with me while I try to explain this crash, let me begin by
  saying I'm using the very first mp sdk source with no updates... and debug
  build is not helping at all. Perhaps the most frightening aspect of this
  crash is its consistancy between game launches, and its inconsistancy
  between builds, its almost as if my compiler is messed up. Basically I'm
  encountering a crash in the client dll. When I create a server, disconnect,
  and then recreate. the debug build gives me an error in
 
  void CBaseViewport::RemoveAllPanels(void) (vgui_teamfortressviewport.cpp
  line 413)
 
  this is the callstack..
 
 client.dll!CBaseViewport::RemoveAllPanels()  Line 413 + 0xf C++
 client.dll!CBaseViewport::~CBaseViewport()  Line 432C++
 client.dll!SDKViewport::~SDKViewport()  + 0x14  C++
 client.dll!SDKViewport::`scalar deleting destructor'()  + 0x14  C++
 client.dll!ClientModeShared::~ClientModeShared()  Line 107 + 0x22
 C++
 client.dll!ClientModeSDKNormal::~ClientModeSDKNormal()  Line 92 + 
  0x8   C++
 client.dll!$E11()  + 0x13   C++
 client.dll!doexit(int code=0, int quick=0, int retcaller=1)  Line 
  376   C
 client.dll!_cexit()  Line 318 + 0xb C
 client.dll!_CRT_INIT(void * hDllHandle=0x2400, unsigned long
  dwReason=0, void * lpreserved=0x0001)  Line 193 C
 client.dll!_DllMainCRTStartup(void * hDllHandle=0x2400, unsigned 
  long
  dwReason=0, void * lpreserved=0x0001)  Line 312 + 0x11  C
 
  The kicker is the call stack indicates the call to this method is happening
  after a call to doExit() which is totally unrelated to anything i've been
  doing - if thats not enough for you, it gets a whole lot worse. I ported one
  single npc over nothing more, and magically the crash was fixed. Now I can
  create/shutdown servers as many times as I want without crashing
 
  It gets worse.
 
  Now when creating an npc I experience the exact same crash. then maybe 30
  rebuilds later.. the crash is magically gone, without even going near npc or
  combat character code of anykind... another few builds later, the crash
  reappears. (totally random)
 
  Keep in mind the crashing is _consistant_ in a build, as in once the crash
  is present no matter how many times i launch the game I'll have the exact
  same behaviour, magically sometimes if I do a rebuild without changing any
  code, the crash vanishes. I checked this to make sure it wasn't something
  like accessing an invalid pointer.
 
  I wasn't even going to post about it here but I thought I'd give it a shot.
  This must be something compiler related (mvsc++7.0) OR for some messed up
  reason I couldn't understand the debugger is showing me the same (old and
  unrelated) callstack, however I don't see how this could be the case as the
  debugger shows the proper information if I crash a different way.
 
  Does *anyone* have *any* suggestions? am I missing some service pack I don't
  know about?
  Thanks in advance...
 
  _
  Powerful Parental Controls Let your child discover the best the Internet has
  to offer.
  http://join.msn.com/?pgmarket=en-capage=byoa/premxAPID=1994DI=1034SU=http://hotmail.com/encaHL=Market_MSNIS_Taglines
   Start enjoying all the benefits of MSN® Premium right now and get the
  first two months FREE*.
 
  ___
  To unsubscribe, edit your list preferences, or view the list archives, 
  please visit:
  http://list.valvesoftware.com/mailman/listinfo/hlcoders
 
 

 ___
 To unsubscribe, edit your list preferences, or view the list archives, please 
 visit:
 http://list.valvesoftware.com/mailman/listinfo/hlcoders




--
spooq

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlcoders



Re: [hlcoders] Bizzare/Random/Frightening crash

2005-02-21 Thread Maurino Berry
unfortunately cleaning it doesnt seem to help very often : I think I'm
going to try compiling on another computer maybe if I can't get any more
suggestions from here.. thanks though :D
_
Designer Mail isn't just fun to send, it's fun to receive. Use special
stationery, fonts and colors.
http://join.msn.com/?pgmarket=en-capage=byoa/premxAPID=1994DI=1034SU=http://hotmail.com/encaHL=Market_MSNIS_Taglines
 Start enjoying all the benefits of MSN® Premium right now and get the
first two months FREE*.
___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlcoders