jis@mtgzx.att.com (J. Mukerji) (02/26/90)
I have noticed a few instances of strange behavior when using Org
Charts. I was wondering if they are generic problems or am I doing
something wrong.
(1) I started with an Org Chart that is created by messages using Expose
Tree. I saved this Org Chart in a file test.org. Then I did a
ez test.org
Ez came up fine and I could do everything with the Org Chart that I
expected to be able to do. However when I tried to kill the buffer
(Delete buffer ^Xk) containing test.org runapp silently dumped core and
ez disappeared.
(2) I started up sendmessage, and I included test.org in the message
using ^X^I. It got included fine. But when I tried to do anything with
the Org Chart in the sendmessage window, sendmessage froze, and then it
managed to crash the X server!!!
(3) If I carefully include an Org Chart in a message and just mail it
without touching it at all then it does indeed go through and get
received by the recipient. But it is a real time bomb. When the
recipient tries to view this message, s/he can view it, but more often
than not, that is the last message that s/he will view with that
pariticular incarnation of messages, because soon thereafter it will
hang with some segmentation violation message displayed on the console.
Has anyone observed these strangenesses? If so, is someone working on
fixing them?
I believe we have patch level 3 installed here. I am not quite sure but
Charlie Hayden <cch@mtgzx.att.com> can confirm or deny that.
Jishnu Mukerji,
jis@mtgzx.att.com,
+1 201 957 5986,
AT&T Bell Laboratories,
MT 3K-423, 200 Laurel Ave.,
Middletown NJ 07748nsb@THUMPER.BELLCORE.COM (Nathaniel Borenstein) (02/26/90)
Excerpts from internet.info-andrew: 25-Feb-90 Org Chart peculiarities J. Mukerji@mtgzx.att.com (1597+0) > I have noticed a few instances of strange behavior when using Org > Charts. I was wondering if they are generic problems or am I doing > something wrong. It is (unfortunately) the case that a bug in any inset can be a time bomb for a program like ez or messages. Messages is particularly vulnerable because when you go on to the next message, it actually destroys the objects in the old message (ez tends to keep them around in a buffer unless you use ^X^K, which is how you got your ez bug). Often the developer of an inset tests how it runs quite carefully, but doesn't so carefully test how it dies, which yields an inset that works fine for demo purposes but causes particular problems when sent through the mail. I conjecture that one of three things is happening in the xxx_FinalizeObject class procedure (where xxx is some one of the many objects in the family of objects that make up the org inset and its apt superclasses). Either something is being freed that shouldn't be, or something is being written past its allocated area, or an enqueued timer event is not being cancelled. I make this conjecture not based on the code (because I don't know the org code, and don't particularly care to) but based on considerably past experience with this type of bug report. I suspect that this hypothesis explains your bugs #1 and #3. As to #2 -- where an org inset in the sendmessage window caused the X server to dump core -- this is obviously harder to diagnose. Clearly there is a bug in the X server, but (if this is reproducible) undoubtedly some ATK/org bug triggers it. Conceivably the symptom will go away when the other bug is fixed, but i wouldn't count on it. -- Nathaniel
gk5g+@ANDREW.CMU.EDU (Gary Keim) (02/27/90)
Excerpts from misc: 25-Feb-90 Org Chart peculiarities J. Mukerji@mtgzx.att.com (1597+0) > I have noticed a few instances of strange behavior when using Org > Charts. I was wondering if they are generic problems or am I doing > something wrong. > Has anyone observed these strangenesses? If so, is someone working on > fixing them? > I believe we have patch level 3 installed here. Your are not doing anything wrong. Do not adjust your set. There have been reports received concering Org but not any that can be reproduced with the hardware prevelent here (rt, 4.3bsd). What machine/os are you using? There is a problem saving org charts on sun3/50s w/ SunOS 4.01. Gary Keim ATK Group
janssen@parc.xerox.com (Bill Janssen) (02/27/90)
Excerpts from andrew: 26-Feb-90 Re: Org Chart peculiarities Nathaniel Borenstein@thu (1800+0) > I conjecture that one of three things is happening in the > xxx_FinalizeObject class procedure (where xxx is some one of the many > objects in the family of objects that make up the org inset and its apt > superclasses). Either something is being freed that shouldn't be, or > something is being written past its allocated area, or an enqueued timer event is not being cancelled. In particular, the suite_FinalizeObject method core-dumps on Sun-4's when called from bushv. (ever wonder why bush takes so long to quit? Busy core-dumping.). And since suite and orgv are both sub-classes of aptv, I'd suspect something in the aptv code. But it's too simple. Looking at orgv_FinalizeObject, we find that it calls treev_Destroy and suite_Destroy. Looking at suite_Destroy, we find it does a number of frees and destroys (but all seem to work when called from orgv_FinalizeObject!). I'll post a note if I find anything. Bill