August 2006 - Darwin-dev - lists.apple.com

Re: Intel XNU bug report
by Chad Leigh 09 Aug '06

09 Aug '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Aug 8, 2006, at 9:09 PM, Eric Albert wrote: On Aug 8, 2006, at 1:55 PM, almisr wrote: But IA-64 is Itanium in most people's mind, that I agree with. Chad -Eric _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… Mike has implemented ACPI 2 standard in IA-64 for Darwin, By the way, IA-64 is the name Intel uses for Itanium, which Darwin does not support. We've been using "Intel 64-bit" or "x86-64" for our latest architecture. Sun I think uses x64 which is kind of a nice shorthand for x86-64. A lot of people like FreeBSD use amd64 since AMD fathered the particular 64bit architecture which intel copied. Would be nice if Darwin/Apple etc used a name that is already in use by others so a to not offer more confusion. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/chad%2Bdarwin% 40objectwerks.com This email sent to chad+darwin(a)objectwerks.com This email sent to site_archiver(a)lists.apple.com

1 0

Re: Ordering of static destructors on exit
by Steve Checkoway 09 Aug '06

09 Aug '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Aug 8, 2006, at 11:14 PM, Rob Cas wrote: I was wondering if anybody can tell me precisely when the destructors of static objects are called. -- Steve Checkoway _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com C++ doesn't not specify the order in which dtors are called. You cannot rely on this order as it can and has changed before. You likely need to restructure your code such that this is not an issue. What I've found to work best is to ensure that dtors of static objects do nothing nontrivial. smime.p7s

1 0

Ordering of static destructors on exit
by Rob Cas 09 Aug '06

09 Aug '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition; b=S5LQxGk4aDz3gMgChLeYKlncnFlrobMX6nYoyzeWT9NCR9gEcQ93SeMHywRgm/uEN2f5FSKyQkDIwohSj1GUm0h95YUm9hTNSpsRt2f97Q8PiKaHN440+UiTWian+mdWBwCkvsCWRpQYOJLSLRFQpthU6iQWzH+/mHfxT/XbDVE= Hi, I was wondering if anybody can tell me precisely when the destructors of static objects are called. I have been using something like: CClass* CClass::singleton() { static CClass singletonInstance; return & singletonInstance; } inside a framework. This framework is used by a QuickTime component. CASE A: In typical applications, the sequence of events I see on quitting the application is: - the QuickTime component closed - destructor of singletonInstance called - application has exited CASE B: In a recent Cocoa application I've come across I'm seeing: - destructor of singletonInstance called - the QuickTime component closed - application has exited (well - application crashes now, as the singleton is referenced by the QuickTime component) I had thought that static destructors always were called at exit() time, so am a little puzzled how the order is reversed by this Cocoa app. (The destructor and component closing appear to be occurring on the same pthread.) Moving the: static CClass singletonInstance; from inside the function, to outside the function (ie. file scope) restores the ordering to that of CASE A, and avoids the crash. But I don't understand why the change of scope would change the ordering of events... Can anybody shed any light, or point me at any only resources that describe how darwin/gcc orders the calling of static destructors? (This may be a cocoa-dev mailing list question, as I understand Cocoa apps are run/controlled by NSApplicationMain() which may do some special atexit() fiddling...but I thought I'd try here first...) Thanks in advance, Rob. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Intel XNU bug report
by Eric Albert 09 Aug '06

09 Aug '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Aug 8, 2006, at 1:55 PM, almisr wrote: -Eric _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… Mike has implemented ACPI 2 standard in IA-64 for Darwin, By the way, IA-64 is the name Intel uses for Itanium, which Darwin does not support. We've been using "Intel 64-bit" or "x86-64" for our latest architecture. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Intel XNU bug report
by almisr 08 Aug '06

08 Aug '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Mike has implemented ACPI 2 standard in IA-64 for Darwin, Mike and Mitsuru and Nate are the ACPI-project on freeBSD so i think if there were such "stupid-code" Mike couldn't miss it :) in freeBSD-IA-64 , ACPI is not implemented yet then Mike is in advance , on fb it gives if (cpu_mode_is64bit()) { void/ apci disable } see also, ACPI interrupt concept for more information http://www.freebsd.org/projects/acpi/ http://freebsd.active-venture.com/handbook/acpi-debug.html Best Regards, {Blocked MAn} {|} On Aug 8, 2006, at 12:22 PM, darwin-dev-request(a)lists.apple.com wrote: Now that the source is released, we might as well get to work. That's the point of releasing the source in the first place, right? File: xnu-792.10.96/osfmk/i386/acpi.c Line: 151 if (cpu_mode_is64bit()) { cpu_IA32e_enable(current_cpu_datap()); cpu_desc_load64(current_cpu_datap()); kprintf("acpi_sleep_kernel 64-bit mode re-enabled\n"); fast_syscall_init64(); } else { fast_syscall_init() } Bug: The 64 bit mode should be restored if the cpu is *NOT* currently in 64 bit mode. The first line should read: if (!cpu_mode_is64bit()) { The diagnostic there is perhaps confusing the matter. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… Without double-checking the source, I think you are making an erroneous assumption here. I believe the applicable logic is that, returning from sleep the CPU will either be in 32-bit mode on a 32-bit system, or 64-bit mode on a 64-bit system. In the former case, nothing need be done as that is the mode in which the kernel operates. On non-64-bit systems, attempting to restore 64-bit mode operation would be fatal. In the latter case, however, the kernel is run in IA32e mode and thus it's necessary to change. = Mike This email sent to site_archiver(a)lists.apple.com

1 0

Re: Apple Opens Up: Mac OS Forge
by Will Barton 08 Aug '06

08 Aug '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Aug 8, 2006, at 4:40 PM, Will Barton wrote: _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com My primary intent was to make sure these questions get asked. I don't expect everything about something as potentially ambitious as Mac OS Forge to fall into place overnight. I do hope, however, that Mac OS Forge can become what OpenDarwin failed to. Sorry, I meant to add the obvious hope that follows: that the problems that caused OpenDarwin to fail are resolved. -Will PGP.sig

1 0

Re: Apple Opens Up: Mac OS Forge
by Will Barton 08 Aug '06

08 Aug '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi Ernie, -Will On Aug 8, 2006, at 1:12 PM, Ernest Prabhakar wrote: Hi Will, Best, -- Ernie P. On Aug 8, 2006, at 8:33 AM, Will Barton wrote: On Aug 7, 2006, at 7:15 PM, Ernest Prabhakar wrote: B. New "Mac OS Forge" for Community Projects -Will _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com Thanks, I imagine WWDC keeps everyone busy, and I appreciate the immediate needs of projects looking for a home from OpenDarwin. My primary intent was to make sure these questions get asked. I don't expect everything about something as potentially ambitious as Mac OS Forge to fall into place overnight. I do hope, however, that Mac OS Forge can become what OpenDarwin failed to. We're all pretty buried with WWDC, but we definitely plan to provide more information once things quiet down. The urgent priority is making sure we have a home for those projects needing to transition off OpenDarwin, as well as Apple's new initiatives like calendaring. Long term, we're looking to the community to find out what -they- need Mac OS X Forge to be, relative to the many other wonderful hosting options available. Mac OS Forge, a new community site hosted by Apple, is being created to support WebKit and other open source projects focused on Mac OS X, especially those looking to transition from OpenDarwin.org. <http://www.macosforge.org/> I have a couple of questions as to the nature of Mac OS Forge. It's unclear exactly what the relationship of Mac OS Forge is to Apple from the website or this announcement. It is also unclear as to how the mentioned "other open source projects" (presumably, external to Apple) would go about being hosted at Mac OS Forge, if they desired. The second question really goes back to the first in a few ways, namely whether an open source project would have to meet some kind of standard of a focus on Mac OS X, and whether that judgement is made within Apple, by members of the community that Mac OS Forge hopes to attract, or some other process altogether, and what degree of transparency can be expected. Transparency is pretty important in building a healthy community, in my very humble opinion. I recognize that Mac OS Forge is new, and there are probably a lot of related issues that need ironing out, but the website is just a little bit vague on actual details. I do apologize ahead of time if all this information is to be made clear in the near future. As to the rest of the announcements, xnu is definitely welcomed, as is the Apache licensing of Bonjour and Launchd! PGP.sig

1 0

Re: Intel XNU bug report
by Michael Smith 08 Aug '06

08 Aug '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Aug 8, 2006, at 12:22 PM, darwin-dev-request(a)lists.apple.com wrote: Now that the source is released, we might as well get to work. That's the point of releasing the source in the first place, right? File: xnu-792.10.96/osfmk/i386/acpi.c Line: 151 if (cpu_mode_is64bit()) { cpu_IA32e_enable(current_cpu_datap()); cpu_desc_load64(current_cpu_datap()); kprintf("acpi_sleep_kernel 64-bit mode re-enabled\n"); fast_syscall_init64(); } else { fast_syscall_init() } Bug: The 64 bit mode should be restored if the cpu is *NOT* currently in 64 bit mode. The first line should read: if (!cpu_mode_is64bit()) { The diagnostic there is perhaps confusing the matter. = Mike _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… Without double-checking the source, I think you are making an erroneous assumption here. I believe the applicable logic is that, returning from sleep the CPU will either be in 32-bit mode on a 32-bit system, or 64-bit mode on a 64-bit system. In the former case, nothing need be done as that is the mode in which the kernel operates. On non-64-bit systems, attempting to restore 64-bit mode operation would be fatal. In the latter case, however, the kernel is run in IA32e mode and thus it's necessary to change. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Apple Opens Up: Mac OS Forge
by Ernest Prabhakar 08 Aug '06

08 Aug '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi Will, Best, -- Ernie P. On Aug 8, 2006, at 8:33 AM, Will Barton wrote: On Aug 7, 2006, at 7:15 PM, Ernest Prabhakar wrote: B. New "Mac OS Forge" for Community Projects -Will _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… We're all pretty buried with WWDC, but we definitely plan to provide more information once things quiet down. The urgent priority is making sure we have a home for those projects needing to transition off OpenDarwin, as well as Apple's new initiatives like calendaring. Long term, we're looking to the community to find out what -they- need Mac OS X Forge to be, relative to the many other wonderful hosting options available. Mac OS Forge, a new community site hosted by Apple, is being created to support WebKit and other open source projects focused on Mac OS X, especially those looking to transition from OpenDarwin.org. <http://www.macosforge.org/> I have a couple of questions as to the nature of Mac OS Forge. It's unclear exactly what the relationship of Mac OS Forge is to Apple from the website or this announcement. It is also unclear as to how the mentioned "other open source projects" (presumably, external to Apple) would go about being hosted at Mac OS Forge, if they desired. The second question really goes back to the first in a few ways, namely whether an open source project would have to meet some kind of standard of a focus on Mac OS X, and whether that judgement is made within Apple, by members of the community that Mac OS Forge hopes to attract, or some other process altogether, and what degree of transparency can be expected. Transparency is pretty important in building a healthy community, in my very humble opinion. I recognize that Mac OS Forge is new, and there are probably a lot of related issues that need ironing out, but the website is just a little bit vague on actual details. I do apologize ahead of time if all this information is to be made clear in the near future. As to the rest of the announcements, xnu is definitely welcomed, as is the Apache licensing of Bonjour and Launchd! This email sent to site_archiver(a)lists.apple.com

1 0

Intel XNU bug report
by Ernesto Corvi 08 Aug '06

08 Aug '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Now that the source is released, we might as well get to work. That's the point of releasing the source in the first place, right? File: xnu-792.10.96/osfmk/i386/acpi.c Line: 151 if (cpu_mode_is64bit()) { cpu_IA32e_enable(current_cpu_datap()); cpu_desc_load64(current_cpu_datap()); kprintf("acpi_sleep_kernel 64-bit mode re-enabled\n"); fast_syscall_init64(); } else { fast_syscall_init() } The first line should read: if (!cpu_mode_is64bit()) { _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… Bug: The 64 bit mode should be restored if the cpu is *NOT* currently in 64 bit mode. Somebody please file a RADAR for this, as I'm having problems filing the report. Thanks, Ernesto. This email sent to site_archiver(a)lists.apple.com

1 0