[Tcl-bugs] [ tcl-Bugs-439941 ] RememberSyncObject() loses the heap

[<no...@so...>]

Bugs item #439941, was opened at 2001-07-09 20:54
You can respond by visiting: 
http://sourceforge.net/tracker/?func=detail&atid=110894&aid=439941&group_id=10894

Category: Embedding Support
Group: 8.4a3
>Status: Closed
>Resolution: Invalid
Priority: 5
Submitted By: David Gravereaux (davygrvy)
Assigned to: Nobody/Anonymous (nobody)
Summary: RememberSyncObject() loses the heap

Initial Comment:
WinNT-5.

Embedding Tcl into a windows GUI application has a 
problem with Tcl_Exit().  Tcl_Finalize() is called 
twice.  Once in Tcl_Exit() and a second time during 
the unloading of tcl84.dll from its DllMain().  In the 
second call to Tcl_Finalize(), TclpInitLock() tries to 
add a new syncobject to keyRecord.list[], but the 
allocation of that once valid memory location has 
disappeared causing the "Invalid memory access" 
dialogbox of death to popup.

I don't have a clear solution to this problem at this 
time.  I would prefer a different codepath for my own 
exit, but I don't see a direct way of setting one, 
like how you can set an app-specific Tcl_PanicProc.  A 
Tcl_SetExitProc() would be a handy API to add.  I 
would use it!  I know for a fact, if I can post a 
WM_CLOSE to the GUI, everything comes down fine.  I 
would prefer not to use ExitProccess() [that's what 
exit() from msvcrt is doing on the inside, btw].

I guess I could write an extension command for 
[exit]...  Is it a valid concern about it getting it 
rename'd at the Tcl level and break what I'm trying to 
fix?

int
MyExit(ClientData c, Tcl_Interp *i, int obj c....)
{
    PostMessage(mainHWnd, WM_CLOSE, 0, 0);
    return TCL_OK;
}
Tcl_CreateObjCommand(interp, "::exit", MyExit, 0L, 0L);

Will that work Ok or do parts of the core call Tcl_Exit
() directly?  I must be able to have the current 
interpreter execution unwind fully so I can delete 
that interpreter, but to do it, I'll have to have 
[exit] return to unwind the current Tcl_Eval().

----------------------------------------------------------------------

Comment By: David Gravereaux (davygrvy)
Date: 2001-07-10 16:27

Message:
Logged In: YES 
user_id=7549

Please mark this invalid.  It was fixed March 30th: Patch 
#403532, Bug #219391

My mistake :)

----------------------------------------------------------------------

You can respond by visiting: 
http://sourceforge.net/tracker/?func=detail&atid=110894&aid=439941&group_id=10894