Dec 4 2004 security update

#1 Updating to 1.5 no help, reverting altivec and g4 tuning no help, I get, pretty consistently, the following crash log.

Since I know bugger all about the mac other than how to push "build" in xcode, I thought I would be a n00b and paste it :)

Any hints would be appreciated.

----
Date/Time: 2004-12-04 14:35:31 -0500
OS Version: 10.3.6 (Build 7R28)
Report Version: 2

Command: Dark Horizons - Lore
Path: /lore/Dark Horizons - Lore.app/Contents/MacOS/Dark Horizons - Lore
Version:

#2 Well just like in any other operating systems debug it on the fly by stepping through the code and checking the values that get passed down the stack trace.

Might be a bug in their security fix that nobody has noticed. So do a debug compile of Torque and all of it's component libraries, and use a visual tool that uses gdb (or something better optimized for the mac OS X platform). It looks like "4 ??? 0x0024639c StartThreadFunc(void*) + 0x34" frame might be responsible, unless of course it screwed up due to the calling function passing it bad data therefore you work yourway up the stack tree.

Also turn on core dumping with unlimited size that way you'll see the whole picture when everything went poof! :)

#3 I was hoping other people who had macs might have seen this ;)

There's actually about 2 more pages to the crash log.

If it is a new security fix and TGE compiles, we'll probably have to conquer this as a group ;)

If it is just me, I am gonna have to find someone to help me fix it ;)

If it's the patch, someone with a mac will have to show me some kindness and tell me how to uninstall the patch :)

#4 Have you tried a google search to see if anybody else has posted about this in a mailing list or something? Might be a good start. :)

#5 I thought that TGE mac developers would be a better place to start, since Lore is TGE ;) I think all google references to mac, and tge and compile problems come back here, and it started after I applied the OS update. Pretty certain this is the right venue for the question (and potential warning). If people haven't applied said OS update (it's actually dated 2 Dec, not 4 Dec), I'd advise against it unless you feel lucky :)

Here's the question more succinctly: Do you have TGE and a Mac? Have you applied this security update? What was the result?


I am getting some expert help, but this all came at a bad time as we wanted to get another beta of Lore Gold out this weekend. Looks like it might have "updated" some frameworks or some permissions to them. Or something. Would be interested in any GG response as well.

#6 Well my research has turned up a lot of people running into the same problem as you can see from google.
Also this might be an interesting try.

That is what I mean by google searching, something tells me you aren't much of a developer are you? :P

#7 Do you even have a mac?

#8 @Edward--you have a lot more politeness/patience than I :) Guess he's not quite sure what Lore is...what do you think, maybe he should google it?

#9 Hi Edward. i too have installed the patch [today], but have not yet tried compiling Torque. i am an extreme novice, but since you were a lone voce i wanted to at least try to help. in grand triage fashion, reverting the patch may be the best bet given the crunch. [i'm waiting for Gold to get my own Lore copy! and it's great to know it's Made on a Mac. at least in part.] the question is figuring out how.

if you are a registered member with Apple Developer Connection, [i am not] they may well have focused forums there. but there is also a public troubleshooting forum which i know is at least monitored by Apple employees, and for that matter it has an extremely active 'bystander' audience.

this thread might or might not relate; he recommends Repair Permissions for a fix for someone's Safari crashing.

http://discussions.info.apple.com/webx?14@19.olQOa4unDkW.1@.68a08f90
"Topic: Security Update 12-02/Safari does not boot up"

[topic title if you need to search.]

then, there is an 'Installation and Setup' forum which seems to cover software updates...

http://discussions.info.apple.com/webx?14@19.olQOa4unDkW.0@.599b3149

but probably your best bet is the [thankfully public] Developer troubleshooting forum: they will speak your language even if they do not have Torque.

http://discussions.info.apple.com/webx?13@19.olQOa4unDkW.0@.599cb753

Forum:
Home > Mac OS X > Mac OS X v10.3 Panther > Developer

i hope these help to some degree, even though it means re-asking your question. at the very least, on any of these boards, *some* sort of answer shld come within minutes; esp the Dev board.

good luck!


- aqaraza

#10 It's something, I have been poking around on ADC all day. Been combing thru my libraries, trying to see what has changed, have a list, some of the core stuff that gets linked to seems to have been updated.

I would be inteerested if you try to compile and tell me what happens. part of me hopes it is a problem with the patch, but that could be bad ;)

If it's just me, well, at least I might be able to get someone else to do my compiles till I restore my machine :)

#11 Just read that first link...

My god, prefs hell, reminds me of 0S7 and 8...

Guess that's alot of system maint to try out while I wait, can't hurt...

#12 I haven't downloaded the security update yet because I read of such issues. I will test it out later this week or early next, I'm heading off for a trip and won't have access to my Mac until then. Thanks to all for helping each other figure this one out in the mean time.

#13 It's been a maddening couple of days.

I think i need to back up all my apps and start over with a fresh machine.

Mark Grob has been helping me out, he's managed to resolve it locally, but my machien still refuses to cooperate.

Happy, happy, joy, joy.

#14 We've opened an Apple incident, it's looking like something in the way the core framworks are accessed.

It's odd that it works on some machines, but not others.

#15 Erased the machine, reinstalled xcode 1.2, updated to OX 10.3.5 and redownloaded my code from CVS.

Now I can compile and run (except for an aggravating little openal linking problem, but i know how to fix that methinks).

Go figure.

#16 @Edward, Glad you made it.

I've been lurking this thread since it started without being able to help. But have someone else reproduced the security update and compile problem ?

I downloaded it but haven't installed ( sounds like a Bill Clinton phrase ).

#17 Update:

There is a problem with RTTI and GCC 3.3 on the mac.

We didn't see it before because of a build setting that suppresses the error with a default value instead of null. When I updated, this setting seems to have been reversed, either stupidly because of my own fat fingering with the compiler, or by the patch, or the patch updated GCC (which can also be update manually). I vote for stupidly by me, but that is neither here, nor there. Don;t know, don;t care, but it does explain the sudden apearance.

The flag is Enable C++ Runtime Types, which gets enabled by default in "development" type builds in Xcode, which made this problem infuriating to chase.

So, it worked before because that setting was checked by default, and stopped when it was mysteriously unchecked. You can alos toggle this setting by switching between "Development" and "Deployment" build settings.

The mac tech had this to say

Quote:The crash looks like an attempt to do a dynamic_cast on an object that contains no runtime type information. To confirm this, I wrote a simple test tool that does a dynamic_cast. By default, the build setting Enable C++ Runtime Types in enabled in the target, so the first run of the test tool completed without problems. I then turned off the Enable C++ Runtime Types option and rebuilt the target. This time, the test tool crashed in the exact same place as your code did.

The workaround is to look for dynamic_casts in the function SceneGraph::addObjectToScene and then build all code that creates instances of that class with Enable C++ Runtime Types enabled.

I have filed against gcc 3.3 since I don't believe a crash is the best way to deal with missing RTTI.

So it will happen with ALL dynamic casts without run time info on the Mac.

Lore is pseudo TGE 1.2 (It's mostly 1.2, for various reasons, we have an incomplete merge, plus all our custom code), and this problem does not appear to exist on post 1.2 TGE. So anyone with 1.3 is safe from the beastie as confirmed by Lightwave Dave.

Special thanks to Lightwave Dave Wyland for helping me fight thru Xcode, he actually found the flag about 10 minutes before the apple tech sent the email. Thanks to Mark Grob for putting up a dts incident (Mark helped us in the last stages of Lore release on the Mac back in April, and is generally a cool guy)

Now you know, and knowing is half the battle.