Path: utzoo!attcan!uunet!aplcen!samsung!zaphod.mps.ohio-state.edu!sdd.hp.com!apollo!caip.rutgers.edu!rutgers!cs.utexas.edu!sdd.hp.com!decwrl!lll-winken!sun-barr!ccut!kogwy!new1!roger
From: roger@zuken.co.jp (Roger Meunier)
Newsgroups: comp.windows.x.motif
Subject: deep in the heap
Message-ID: <4c0e7f10.20b6d@apollo.HP.COM>
Date: 7 Aug 90 20:26:00 GMT
Sender: root@apollo.HP.COM
Organization: ZUKEN Inc. Yokohama, JAPAN
Lines: 37
To: comp-windows-x-motif@rutgers.edu


A few weeks ago, I got my hands on Conor P. Cahill's (uunet!virtech!cpcahil)
malloc/string debugging library.  I linked it with one of my Motif-based
test applications to check for memory leaks, etc. in *my* code but ran into
problems with the Motif library first.  The memory-leak issue has been
dealt with before, so I won't rehash it here.  What concerns me is that
the *bugs* I encountered could have been easily caught if someone testing
Motif *before* letting it out into the real world would have done the simple
tests which Mr. Cahill's library allows.  For example:

	1)  Accessing free()'d memory.  I hope this one is fixed in the
	next release.  It occurred with my application and also with
	xmdialogs (from xmsamplers).  Destroying any widget that contains
	a command, file selection box/dialog, or selection box/dialog
	and then returning to the event loop results in sombody trying
	to access the XmScrollBar widget *after* it has been deallocated.
	(Mr. Cahill's tool fills deallocated memory with a pattern which
	causes most applications to choke if they try to access it again.
	Motif choked at _XmAtomtoName+0x12a.)

	2)  Overrunning the end of an allocated buffer.  I tried copying
	a string between XmText widgets using the secondary buffer.
	Motif apparently does a realloc() on some internal buffer, at
	which time it was discovered that someone had written 6 bytes
	into the buffer when only 5 were allocated.  (On my machine, this
	didn't cause the program to crash, but I do remember a few
	months ago when my program *did* crash when I used this feature,
	but at that time I blamed it on my own coding.  Who knows what
	else is lurking here...)

Seeing as there are tools available which catch these sort of problems,
I can only hope that they become widely used in the future.  When
trying to shake bugs out of my own code, it would be nice if I didn't
have to hassle with someone elses... now who wants to argue with THAT?

(I am using Motif 1.0 with X11R3 on HP9000 300-series under HP-UX 7.0.)
--
Roger Meunier @ Zuken, Inc.  Yokohama, Japan	(roger@zuken.co.jp)