1109 replies to this topic

[BlackCH]

DB1, on Jan 25 2010, 01:38 PM, said:

@BlackCH - would you be so kind to make the P5K VM - bios rom with the attached dsdt.dsl so I can measure the difference between loading an updated dsdt from bios as opposed to via chameleon.

Sure mate!; I have work piled up but I'll try to do it ASAP

[Master Chief]

MacKonsti, on Jan 25 2010, 10:45 AM, said:

Guys, I have added the following kexts/services in my mobo, that could slow down the process. Judging from the VMWare removal reported by Master Chief, please comment...

(1) lspci.kext (64-bit)
(2) little snitch service (appears in the boot log, too)

Number 2 is taking its time yes.

MacKonsti, on Jan 25 2010, 10:45 AM, said:

As I mentioned in my previous post (which nobody commented) is there a serious difference in detecting/loading method of the ethernet card between (a) injected plist kext via chameleon, and ( system native detection (by addind h/w ID in the plist of the driver)?

No.  That is basically the same but...

Chameleon loads Extensions.mkext ahead of time and that is why you got the warning(s).  The good news is that we can eliminate this process.  We don't really need it anymore.  More about this at a later stage, but the clue lies in adding symbolic links to our legacy kexts.

p.s. Sometimes people are busy with other stuff and simply don't have the time to reply.  Nothing personal.

[MacKonsti]

Master Chief, I also forgot VoodooMonitor, so now we have these three non-vanilla kexts:

(1) lspci.kext (64-bit)
(2) LittleSnitch 2.2.05 service
(3) VoodooMonitor

that could be delaying boot times. While verbose, LittleSnitch doesn't seem to slow down the boot process that much, it's fast on-screen. Do you really thing it's a "sluggie"? (from slug+proggie)

Also, I get a message: mbinit: done (64 MB memory set for mbuf pool) any idea what this is?

Finally, should there be any delays between booting at 32-bit mode and 64-bit mode? Just to clarify this and take it out of the equation...

Master Chief, on Jan 25 2010, 05:27 PM, said:

Chameleon loads Extensions.mkext ahead of time and that is why you got the warning(s).  The good news is that we can eliminate this process.  We don't really need it anymore.  More about this at a later stage, but the clue lies in adding symbolic links to our legacy kexts.

I am duying to know more

Also, when booting with a DVD-R (burned, not blank) in my SATA optical drive, I get this message/error: disk2: ioctl(_IOWR,'d',132,24) is unsupported. So when measuring boot times, we must make sure it's done without optical discs inside...

Master Chief, on Jan 25 2010, 05:27 PM, said:

p.s. Sometimes people are busy with other stuff and simply don't have the time to reply.  Nothing personal.

I am aware it's not personal; I was just worried that I was under the wrong impression as to what this thread wants to achieve and that I was posting irrelevant info/ideas... Thanks for comforting us. Good night from Athens

[scrax]

I've just finished patching my bootloader to remove hpet and change the spinner, now let's see if it boot up

Ok it's working now without hpet but still same number of spins.
I've managed to change the spinner before the apple logo with letters but I cant compile without errors if use also this code on the next line after the spinner:

static char		  string[2] = {'\0', '\0'};

For now I can count till F the first spin at the top of the screen and till D in the second one few line under the first.
I think that now it's time to try the unused kext remotion.

[rednous]

Master Chief, very nice and advanced guide

I would like to know more about the softwares needed to complete the above advanced tasks. i will try with my BIOS, definitely.

Just for the record to mention mine OS X boot time details (spinner revolutions):

LAN & Audio enabled in BIOS: 15 revolutions
LAN & Audio disabled in BIOS: 6,5 revolutions (1st boot), 5,5 revolutions (2nd boot)

[scrax]

This is a little reminder for me and to make more clear the situation to all of you.

SSD Drive(s) : No
Removing Unused Kexts : Not yet
DSDT Patching / Stripping : Ok
DSDT Loading : Revolution
Incorporating DSDT : Yes
Incorporating SSDT Table : No
Gathering Data : Ok
Text Spinner : Ok first: F , second: D
Boot Logo : 9 to 10 spin
Log files link
A First Test
I'm using Revolution on a P5KC with P5KR bios to enable ICH9 AHCI, on my board I can disable Audio (A), Lan (L), Jmicron Pata/Sata (J) and Firewire (F). I've made 3 or more reboot for each test at 32bit and the results are (in spins under the apple logo):
N1: 9,5 - 10 - 9
A: 8 - 8,5 - 8
L: 11 - 9,5 - 10
J: 9 - 9 - 9,5
F: 8,5 - 9,5 - 10,5 - 9,5
All:9 - 9 - 9 - 9,5
Verifying HPET compatibility : Ok (removed all suggestion)
Preparing for dsdt_patcher removal : Yes (already in Revolution 0.4)

SSDT:
 SSDTv9.dsl.zip   2.01K   205 downloads

Note: No EFI string

[waldo_pepperjack]

Master Chief, on Jan 24 2010, 09:18 AM, said:

Work in progress – Postponed Until?; one of my children is hospitalized (victim of a hit and run accident).

We all wish your child, you, and your family the best.

[rednous]

Master Chief,

I believe everything's gonna be OK with your child. Wish you and your family all the best and good luck.

[Master Chief]

MacKonsti, on Jan 26 2010, 12:42 AM, said:

Master Chief, I also forgot VoodooMonitor, so now we have these three non-vanilla kexts:

(1) lspci.kext (64-bit)
(2) LittleSnitch 2.2.05 service
(3) VoodooMonitor

that could be delaying boot times. While verbose, LittleSnitch doesn't seem to slow down the boot process that much, it's fast on-screen. Do you really thing it's a "sluggie"? (from slug+proggie)

Try to boot without the kexts and see what it does.  I removed it because it did slow down the boot process yes.

MacKonsti, on Jan 26 2010, 12:42 AM, said:

Also, I get a message: mbinit: done (64 MB memory set for mbuf pool) any idea what this is?

Network memory pool (see: netstat -a).

MacKonsti, on Jan 26 2010, 12:42 AM, said:

Finally, should there be any delays between booting at 32-bit mode and 64-bit mode? Just to clarify this and take it out of the equation...

Here 64 bit takes more time and used more memory so it's not just speed what counts.

Edit: Thanks for the kind words folks.  We're not there yet.  Might take months to recover.  A long way ahead of us so to speak.

[scrax]

scrax, on Jan 26 2010, 09:26 AM, said:

I've managed to change the spinner before the apple logo with letters but I cant compile without errors if use also this code on the next line after the spinner:

static char		  string[2] = {'\0', '\0'};

I've found what to do to compile right there is a line static char      string [3] = that needs to be changed.
Now it's ok.

i've also removed in this order from Revolution v0.4 :
ext2fs.o ntfs.o msdos.o commenting some lines in disc.c
usb.o editing ehci & uhci line in pci_setup.c
ati.o & nvidia.o editing pci_setup.c
ufs.o ufs_byteorder.o \ nothing to change
pci_setup.o comment in fake_efi.c:

// extern void setup_pci_devs(pci_dt_t *pci_dt);
/code]
and
[code]/*void setupEfiDevices(void)
{
	setup_pci_devs(root_pci_dev);
}*/

and

//	setupEfiDevices();

nbp.o
/boot2/boot.c
commented this:

	if ( (gBootFileType == kNetworkDeviceType) && gUnloadPXEOnExit ) {
//		if ( (ret = nbpUnloadBaseCode()) != nbpStatusSuccess )
//		{
			printf("nbpUnloadBaseCode error %d\n", (int) ret);
			sleep(2);
//		}
	}

and this:

//	nbpUnloadBaseCode();

saio_internal.c commented this line:

/* nbp.c */
// extern UInt32 nbpUnloadBaseCode();
// extern BVRef  nbpScanBootVolumes(int biosdev, int *count);

sys.c commented this:

  if (bvr == NULL)
  {
 /*   bvr = nbpScanBootVolumes(biosdev, count);
	if (bvr != NULL)
	{
	  gBootFileType = kNetworkDeviceType;
	}
  */}

now let's see if it still boots

It boot's ok, but I don't see any difference in boot time.

For what i can understand this was done to have a smaller bootloader, now it is: 147.776 byte instead of: 148.416 byte 26.360 byte less am I on the good way?

Is this the Revolution 0.5 you have in your sign or is the one with the ssdt injected?


EDIT: I'm trying to avoid the use of smbios.plist and so i'm edting the default value in smbios_patcher.c to match what i have. Then i'll try to eliminate the use of c.a.b.p but i need to know how to put my SystemUUID as default, and to make the sleep fix active by default.

[Master Chief]

scrax, on Feb 3 2010, 05:32 AM, said:

...
i've also removed in this order from Revolution v0.4 :
ext2fs.o ntfs.o msdos.o commenting some lines in disc.c
usb.o editing ehci & uhci line in pci_setup.c
ati.o & nvidia.o editing pci_setup.c
ufs.o ufs_byteorder.o \ nothing to change
pci_setup.o comment in fake_efi.c:
...

now let's see if it still boots

Removing objects from Makefile is one thing you have to do yes, but the goal is to change the Chameleon's source code in a way that you can compile the boot2 folder... as if nothing really changed.  So instead of commenting out void setupEfiDevices(void) I changed it like so:

#if CHAMELEON
void setupEFIDevices(void)
{
	setup_pci_devs(root_pci_dev);
}
#endif

Note that I did change the function name, but it's still there.  And instead of doing this for most of the Chameleon features, I used feature specific stuff like: #if RAMDISK_SUPPORT, #if CDBOOT_SUPPORT, #if LANBOOT_SUPPORT, #if HIBERNATE_SUPPORT, #if TEXT_SPINNER and of course GUI_SUPPORT

This way people (like me) can exclude certain features when they don't want / need it.  Much more fun building Chameleon this way.  It does add a layer of complexity, but mostly for junior developers and other people new to this kind of development.  It is however just the logical thing to do.

scrax, on Feb 3 2010, 05:32 AM, said:

It boot's ok, but I don't see any difference in boot time.
For what i can understand this was done to have a smaller bootloader, now it is: 147.776 byte instead of: 148.416 byte 26.360 byte less am I on the good way?

It's a start yes.  And booting from a slower device, like an USB memory stick, will make it painfully clear that Chameleon has speed issues. Remember this; The faster the system the less you gain here.

p.s. Maybe I need my breakfast first, or a head examination, but here 148.416 - 147.776 = 640 and not 26.360  Typo?

scrax, on Feb 3 2010, 05:32 AM, said:

Is this the Revolution 0.5 you have in your sign or is the one with the ssdt injected?

That's Revolution 0.6 (already) with DSDT and SSDT injection (incorporated).

scrax, on Feb 3 2010, 05:32 AM, said:

EDIT: I'm trying to avoid the use of smbios.plist and so i'm edting the default value in smbios_patcher.c to match what i have. Then i'll try to eliminate the use of c.a.b.p but i need to know how to put my SystemUUID as default, and to make the sleep fix active by default.

You can replace the default static EFI_CHAR8 const SYSTEM_ID[] = {...}
with the UUID of your liking for this.  This is also what I did.

BTW: What sleep fix are you talking about?

[scrax]

Master Chief, on Feb 3 2010, 04:54 PM, said:

BTW: What sleep fix are you talking about?

Not sleep, sorry, your Restart FACP table. I use RestartFix=Yes in c.a.B.p . For now i've solved using the system c.a.B.p with all the option i need in Kernel Flags.
I'll try to add the UUID in the code.

Yes dimensions was a typo,
Revolution 0.4.2 (before removing anything) is 147.776 bytes
Revolution 0.4.4 (with the editing) is 121.728 bytes
147.776-121.728=26.048

[Master Chief]

I have something for people to look at:

Feb  3 11:57:29 localhost kernel[0]: npvhash=4095
Feb  3 11:57:29 localhost kernel[0]: PAE enabled
Feb  3 11:57:29 localhost kernel[0]: 64 bit mode enabled
Feb  3 11:57:29 localhost kernel[0]: Darwin Kernel Version 10.2.0: Tue Nov  3 10:37:10 PST 2009; root:xnu-1486.2.11~1/RELEASE_I386
Feb  3 11:57:29 localhost kernel[0]: vm_page_bootstrap: 970449 free pages and 78127 wired pages
Feb  3 11:57:29 localhost kernel[0]: standard timeslicing quantum is 10000 us
Feb  3 11:57:29 localhost kernel[0]: mig_table_max_displ = 73
Feb  3 11:57:29 localhost kernel[0]: AppleACPICPU: ProcessorId=1 LocalApicId=0 Enabled
Feb  3 11:57:29 localhost kernel[0]: AppleACPICPU: ProcessorId=2 LocalApicId=1 Enabled
Feb  3 11:57:29 localhost kernel[0]: AppleACPICPU: ProcessorId=3 LocalApicId=2 Enabled
Feb  3 11:57:29 localhost kernel[0]: AppleACPICPU: ProcessorId=4 LocalApicId=3 Enabled
Feb  3 11:57:29 localhost kernel[0]: calling mpo_policy_init for Quarantine
Feb  3 11:57:29 localhost kernel[0]: Security policy loaded: Quarantine policy (Quarantine)
Feb  3 11:57:29 localhost kernel[0]: calling mpo_policy_init for Sandbox
Feb  3 11:57:29 localhost kernel[0]: Security policy loaded: Seatbelt sandbox policy (Sandbox)
Feb  3 11:57:29 localhost kernel[0]: calling mpo_policy_init for TMSafetyNet
Feb  3 11:57:29 localhost kernel[0]: Security policy loaded: Safety net for Time Machine (TMSafetyNet)
Feb  3 11:57:29 localhost kernel[0]: Copyright © 1982, 1986, 1989, 1991, 1993
Feb  3 11:57:29 localhost kernel[0]: The Regents of the University of California. All rights reserved.
Feb  3 11:57:29 localhost kernel[0]: MAC Framework successfully initialized
Feb  3 11:57:29 localhost kernel[0]: using 16384 buffer headers and 4096 cluster IO buffer headers
Feb  3 11:57:29 localhost kernel[0]: IOAPIC: Version 0x20 Vectors 64:87
Feb  3 11:57:29 localhost kernel[0]: ACPI: System State [S0 S3 S4 S5] (S3)
Feb  3 11:57:29 localhost kernel[0]: RTC: Only single RAM bank (128 bytes)
Feb  3 11:57:29 localhost kernel[0]: mbinit: done (64 MB memory set for mbuf pool)
Feb  3 11:57:29 localhost kernel[0]: From path: "uuid", 
Feb  3 11:57:29 localhost kernel[0]: Waiting for boot volume with UUID 00000000-0000-0000-0000-000000000000
Feb  3 11:57:29 localhost kernel[0]: Waiting on <dict ID="0"><key>IOProviderClass</key><string ID="1">IOResources</string><key>IOResourceMatch</key><string ID="2">boot-uuid-media</string></dict>
Feb  3 11:57:29 localhost kernel[0]: com.apple.AppleFSCompressionTypeZlib load succeeded
Feb  3 11:57:29 localhost kernel[0]: AppleIntelCPUPowerManagementClient: ready
Feb  3 11:57:29 localhost kernel[0]: Got boot device = IOService:/AppleACPIPlatformExpert/PCI0@0/AppleACPIPCI/SATA@1F,2/AppleAHCI/PRT0@0/IOAHCIDevice@0/AppleAHCIDiskDriver/IOAHCIBlockStorageDevice/IOBlockStorageDriver/SAMSUNG HD103SJ Media/IOGUIDPartitionScheme/Snow Leopard Boot@2
Feb  3 11:57:29 localhost kernel[0]: BSD root: disk1s2, major 14, minor 5
Feb  3 11:57:29 localhost kernel[0]: FireWire (OHCI) VendorID 1106 ID 3044 built-in now active, GUID 0000000000000000; max speed s400.
Feb  3 11:57:29 localhost kernel[0]: systemShutdown false
Feb  3 11:57:31 Chiefs-Mac-Pro kernel[0]: Previous Shutdown Cause: 0
Feb  3 11:57:31 Chiefs-Mac-Pro kernel[0]: NVDANV50HAL loaded and registered.
Feb  3 11:57:31 Chiefs-Mac-Pro kernel[0]: DSMOS has arrived
Feb  3 11:57:31 Chiefs-Mac-Pro kernel[0]: AppleIntelCPUPowerManagement: initialization complete
Feb  3 11:57:33 Chiefs-Mac-Pro kernel[0]: AppleYukon2: Marvell Yukon Gigabit Adapter 88E8056 Singleport Copper SA
Feb  3 11:57:33 Chiefs-Mac-Pro kernel[0]: AppleYukon2: RxRingSize <= 1024, TxRingSize 256, RX_MAX_LE 1024, TX_MAX_LE 768, ST_MAX_LE 3328
Feb  3 11:57:33 Chiefs-Mac-Pro kernel[0]: yukon: Ethernet address 00:00:00:00:00:00
Feb  3 11:57:34 Chiefs-Mac-Pro kernel[0]: Ethernet [AppleYukon2]: Link up on en0, 100-Megabit, Full-duplex, No flow-control, Debug [796d,6c08,0de1,0200,41e1,4000]

Remember this: It took up to 19 seconds when I started and now only 5 seconds.   That to me is a pretty amazing step forward.

I am also pretty sure that implement a hard coded / patched SMBIOS table will help take off another second.  Yeah.  Booting has never been this exiting.   And you know what; My sons brand new quad code iMac is slower.  Ha! Eat that Apple.

scrax, on Feb 3 2010, 11:48 PM, said:

Not sleep, sorry, your Restart FACP table. I use RestartFix=Yes in c.a.B.p . For now i've solved using the system c.a.B.p with all the option i need in Kernel Flags.

The RestartFix=Yes option is never going to work in Revolution because it is pointless to have options for features like this.  Yes.  We want to be able to restart our hacks, but don't need to have such parameter in whatever file they come up with.

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

Edit: Why have all these different locations for stuff like /Extra/ et all when you only need it on your boot drive?  I don't get it.  I mean your hardware surely isn't going to change when you boot from drive 3 instead of 2 or 1?  That would be something wouldn't it.

Yeah yeah.  Now some of you say: But I have 10.5.8 on one drive and 10.6.2 on another.  Which is fine because it loads the stuff it needs from the boot drive.  Whatever drive that is.  But not here, because I only have it on my USB stick.  Much cleaner.  Oh yeah.  You can keep your EFI partition tricks right where they are... in the tool box for people who like to do stuff them selfs :-)

[afiser]

this thread is interesting. just got the hang of DSDT and now something more complicated came out.. time to research! lol

[scrax]

Master Chief, on Feb 4 2010, 12:15 AM, said:

Edit: Why have all these different locations for stuff like /Extra/ et all when you only need it on your boot drive?  I don't get it.  I mean your hardware surely isn't going to change when you boot from drive 3 instead of 2 or 1?  That would be something wouldn't it.

Yeah yeah.  Now some of you say: But I have 10.5.8 on one drive and 10.6.2 on another.  Which is fine because it loads the stuff it needs from the boot drive.  Whatever drive that is.  But not here, because I only have it on my USB stick.  Much cleaner.  Oh yeah.  You can keep your EFI partition tricks right where they are... in the tool box for people who like to do stuff them selfs :-)

I tried to boot from usb and I have a little slower boot up time (12 spin instead of 10 on average) but it's better than i was hoping for.
It's surely more clean and i can have more usb with different bootloader so it's easy to test new version without any risk.

I don't know, because i've not yet tried if I still need to have boot in the root of the volume were is installed SnowLeopard. I think not because boot is already loaded when i chose the volume to boot, but what about boot2 or boot1h?
Another "problem" is I need to install Attansic1LEthernet.kext in S/L/E is there a method of avoid this, maybe using a .mkext with all the .kext i use on the usb?

Now my boot is like this:

USB Bootloader 1:

/boot  (chameleon2RC5pre8)
/com.apple.Boot.plist
/dsdt.aml (ExtremeInsanely P5K Pro thread based)
/smbios.plist
/Extensions/fakeSMC.kext (2,5 debug=false)
/Extensions/LegacyHDAController.kext
/Theme/Nice/

USB Bootloader 2:

/boot  (Revolution 0.4."5" smbios value & my UUID made default in boot)
/Extensions.mkext (fakeSMC.kext, LegacyHDAController.kext)
/ACPI/SSDT.aml (with all the device and id injections and P-State value)

SnowLeo volume:

/S/L/E/Attansic1LEthernet.kext
/S/L/E/AppleHDA.kext (not need, just to speed up the boot)

[Master Chief]

scrax, on Feb 5 2010, 04:44 PM, said:

I tried to boot from usb and I have a little slower boot up time (12 spin instead of 10 on average) but it's better than i was hoping for.
It's surely more clean and i can have more usb with different bootloader so it's easy to test new version without any risk.

What is the difference in load time between USB 1 and 2?

And since we don't do anything with /Extra/Extensions and /Extra/Extensions.mkext anymore... you either have to add the required kext files to /System/Library/Extensions or use symbolic links.  This will change but not in this update.

p.s. I e-mail you a test version of Revolution v0.6 so that you can have a go with it.  Which of course bounced back because gmail won't accept ZIP attachments.  Stupid but true.

[scrax]

Master Chief, on Feb 5 2010, 09:34 PM, said:

What is the difference in load time between USB 1 and 2?

And since we don't do anything with /Extra/Extensions and /Extra/Extensions.mkext anymore... you either have to add the required kext files to /System/Library/Extensions or use symbolic links.  This will change but not in this update.

p.s. I e-mail you a test version of Revolution v0.6 so that you can have a go with it.  Which of course bounced back because gmail won't accept ZIP attachments.  Stupid but true.

I don't know how to boot with usb1, if i disable in bios usb2 i get a lot of error at boot, and so i don't know if it's slow for the slower transfer rate or for the errors...
Now I have all in S/L/E but it seems a little slower just one round more...
Just a note with Revolution v0.4 i need to keep /Extra/ACPI/SSDT.aml on the OsX volume cause it not loads from the usb.

I was looking for the mail, wasn't aware of this problem, with gmail... really stupid. Renaming it archive.zip.jpg doesn't work, does?

[Master Chief]

scrax, on Feb 6 2010, 02:28 PM, said:

I don't know how to boot with usb1, if i disable in bios usb2 i get a lot of error at boot, and so i don't know if it's slow for the slower transfer rate or for the errors...

More work for Rekursor I guess.

scrax, on Feb 6 2010, 02:28 PM, said:

Now I have all in S/L/E but it seems a little slower just one round more...

That is very strange because not having to open Extensions.mkext (and the lack of the extra checks for it) should make it faster.

scrax, on Feb 6 2010, 02:28 PM, said:

Just a note with Revolution v0.4 i need to keep /Extra/ACPI/SSDT.aml on the OsX volume cause it not loads from the usb.

I have incorporated my SSDT.aml so I wouldn't know if this was fixed with any of the later updates, but I will verify this later today.

scrax, on Feb 6 2010, 02:28 PM, said:

I was looking for the mail, wasn't aware of this problem, with gmail... really stupid. Renaming it archive.zip.jpg doesn't work, does?

I tried a few things but no go.  And I won't attach it here simply because I'm not done with it.  Just mail me from a different account when you still want to give it a go.

[scrax]

Master Chief, on Feb 6 2010, 07:31 PM, said:

More work for Rekursor I guess.
I have incorporated my SSDT.aml so I wouldn't know if this was fixed with any of the later updates, but I will verify this later today.

I'm using the sources you posted in the P5K pro thread so it's based on RC4 not on the last RC5, so the SSDT loading is the one you addd not the one rek's is adding to RC5.

I you think it's better to start with a "fresh" chameleon source, maybe we need to use all the same, or not?

It's all day that i have strange results fron my test, actually I cant anymore restart with Revolution but i have sleep and ibernation working, with Rc5pre8 i can't sleep but i can restart...
Time to clean up something...

[Master Chief]

scrax, on Feb 6 2010, 07:48 PM, said:

I'm using the sources you posted in the P5K pro thread so it's based on RC4 not on the last RC5, so the SSDT loading is the one you addd not the one rek's is adding to RC5.

I have e-mailed you Revolution 0.6.15 and today I hope to fix verbose mode with 0.6.16 (boot prompt shows up already, but isn't working yet.  Should be fine later today).  

scrax, on Feb 6 2010, 07:48 PM, said:

I you think it's better to start with a "fresh" chameleon source, maybe we need to use all the same, or not?

I used Chameleon RC4 because it is the latest official release.  And I will keep using this release until something better comes up, or abandon the official releases altogether and only use whatever I need from it (I'm not too happy with some of the patches I see popping up).

Let's take one simple example.  Let's take the safe_malloc patch that went in recently.  Not only enables my implementation you to reduce boot(2) with 768 bytes.  Which might not look like a lot, but I need it for the in-BIOS version of Revolution.  But more importantly: When was the last time you, or anyone else here, have seen this memory allocation error message?  Never I presume.  Which begs the question: Should we even implement it in non-debug versions?  I would say no.