April 2005 - Darwin-dev - lists.apple.com

Re: How do I prevent a function from being optimized out?
by Andrew Gallatin 13 Apr '05

13 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Brian Dantes writes: > No, you do not need to force a flat namespace with mach_override, that's the > beauty of it. It is a bit scary when you realize what it's doing - but hey > you'll learn to live with it. ;-> Now, if you want to override something in Wait, how can that work? Are you saying that even though there are multiple namespaces, the "different" malloc functions wind up at the same address, or do you have to do the override in each namespace? Drew _______________________________________________ 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: How do I prevent a function from being optimized out?
by Brian Dantes 13 Apr '05

13 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Microsoft-Entourage/10.1.4.030702.0 No, you do not need to force a flat namespace with mach_override, that's the beauty of it. It is a bit scary when you realize what it's doing - but hey you'll learn to live with it. ;-> Now, if you want to override something in a program you can't link against (ala the Linux LD_PRELOAD trick), you'll have to wait for Tiger to be able to do that. -Brian > From: Andrew Gallatin <gallatin(a)cs.duke.edu> > Date: Wed, 13 Apr 2005 18:21:50 -0400 (EDT) > To: Brian Dantes <brian(a)things.org> > Cc: <darwin-dev(a)lists.apple.com> > Subject: Re: How do I prevent a function from being optimized out? > > > Brian Dantes writes: >> If you have control over the application, a much better solution is >> mach_override. >> >> <http://www.rentzsch.com/mach_override/> >> > > Actually, that is a good idea. I wasn't aware of this... > All we really need to do is to call a function from > deallocate_pages() with the address and the size. I suppose > we would still need to force a flat namespace. > > If I can't figure anything else out, I may try this. > Testing a totally new way of doing this will not be fun though :( > > Thanks a lot for the pointer! > > Drew > _______________________________________________ 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: How do I prevent a function from being optimized out?
by Andrew Gallatin 13 Apr '05

13 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Brian Dantes writes: > If you have control over the application, a much better solution is > mach_override. > > <http://www.rentzsch.com/mach_override/> > Actually, that is a good idea. I wasn't aware of this... All we really need to do is to call a function from deallocate_pages() with the address and the size. I suppose we would still need to force a flat namespace. If I can't figure anything else out, I may try this. Testing a totally new way of doing this will not be fun though :( Thanks a lot for the pointer! Drew _______________________________________________ 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: How do I prevent a function from being optimized out?
by Andrew Gallatin 13 Apr '05

13 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Jerry Pendergraft writes: > Seems like if you provide an object file to the link that provides > the malloc() entry point (with proper signature), then the linker would > never bother to search the standard lib for it. Of course gcc has its own > ideas especially concering malloc ;^) This is a library itself, and its highly likely somebody else has already called malloc. The first caller initiates the lazy binding search, and the linker pulls in the one from libc. Drew _______________________________________________ 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: How do I prevent a function from being optimized out?
by Jerry Pendergraft 13 Apr '05

13 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Seems like if you provide an object file to the link that provides the malloc() entry point (with proper signature), then the linker would never bother to search the standard lib for it. Of course gcc has its own ideas especially concering malloc ;^) On Wed, 13 Apr 2005, Andrew Gallatin wrote: > > Ben writes: > > I don't know, but the obvious(?) question I have is, "why?" > > > > I was hoping nobdy would ask ;) > > Because we need to override malloc, and the only way we can figure out > to do it is to > > - force a flat namespace using DYLD_FORCE_FLAT_NAMESPACE > - link applications with -multiply_defined -u foo_hack() > - add foo_hack() into scalable_malloc.c along with the hooks we need > - include the scalable_malloc.o object in our lib. > > As I understand it (I didn't come up with it) this forces the rtld to > load our scalable_malloc.o rather than the one from libc, since the first > thing it does is to go looking for foo_hack(). > > Drew > > _______________________________________________ > 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/jerry%40endocardial.com > > This email sent to jerry(a)endocardial.com > _______________________________________________ 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: How do I prevent a function from being optimized out?
by Brian Dantes 13 Apr '05

13 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Microsoft-Entourage/10.1.4.030702.0 If you have control over the application, a much better solution is mach_override. <http://www.rentzsch.com/mach_override/> If you need to override malloc in programs you cannot link against, you will have to wait for Tiger (unless you don't care when the override occurs, then look at mach_inject on the same site). -Brian Dantes > From: Andrew Gallatin <gallatin(a)cs.duke.edu> > Date: Wed, 13 Apr 2005 17:50:06 -0400 (EDT) > To: Ben <bench(a)silentmedia.com> > Cc: darwin-dev(a)lists.apple.com > Subject: Re: How do I prevent a function from being optimized out? > > > Ben writes: >> I don't know, but the obvious(?) question I have is, "why?" >> > > I was hoping nobdy would ask ;) > > Because we need to override malloc, and the only way we can figure out > to do it is to > > - force a flat namespace using DYLD_FORCE_FLAT_NAMESPACE > - link applications with -multiply_defined -u foo_hack() > - add foo_hack() into scalable_malloc.c along with the hooks we need > - include the scalable_malloc.o object in our lib. > > As I understand it (I didn't come up with it) this forces the rtld to > load our scalable_malloc.o rather than the one from libc, since the first > thing it does is to go looking for foo_hack(). > > Drew > > _______________________________________________ > 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/brian%40things.org > > This email sent to brian(a)things.org > _______________________________________________ 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: How do I prevent a function from being optimized out?
by Andrew Gallatin 13 Apr '05

13 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Ben writes: > I don't know, but the obvious(?) question I have is, "why?" > I was hoping nobdy would ask ;) Because we need to override malloc, and the only way we can figure out to do it is to - force a flat namespace using DYLD_FORCE_FLAT_NAMESPACE - link applications with -multiply_defined -u foo_hack() - add foo_hack() into scalable_malloc.c along with the hooks we need - include the scalable_malloc.o object in our lib. As I understand it (I didn't come up with it) this forces the rtld to load our scalable_malloc.o rather than the one from libc, since the first thing it does is to go looking for foo_hack(). Drew _______________________________________________ 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: How do I prevent a function from being optimized out?
by Ben 13 Apr '05

13 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com I don't know, but the obvious(?) question I have is, "why?" On Wed, 13 Apr 2005, Andrew Gallatin wrote: > Is there a supported way to prevent ld from optimizing out an empty > function from a library? Eg: > _______________________________________________ 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

How do I prevent a function from being optimized out?
by Andrew Gallatin 13 Apr '05

13 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi, I'm hoping this is the right list for this question... I'm more of a driver guy. Is there a supported way to prevent ld from optimizing out an empty function from a library? Eg: foo_hack(void) { return; } The best we've been able to do is to make foo_hack() actually do something: extern int dummy_variable; foo_hack(void) { /* do some work to avoid being optimized away */ dummy_variable++; } And then call it in another module: int dummy_variable; some_other_function(..) { <...> /* call a function in our malloc routine so ld doesn't optimize it away */ foo_hack(); <...> } This works on most supported toolchains on MacOSX, but we recently ran into one being used by http://g95.org (ftp://gcc.gnu.org/pub/gcc/infrastructure/cctools-528.5.dmg) where this doesn't work. Is there another, better, way? Thanks, Drew _______________________________________________ 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: Unicode string(3) library?
by Alexey Proskuryakov 13 Apr '05

13 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Wed, 13 Apr 2005 12:29:52 -0500 David Hoerl <dhoerl(a)mac.com> wrote: Does such a thing exist? - WBR, Alexey Proskuryakov _______________________________________________ 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… I am looking for a version of the string(3) library, but adapted for Unicode strings. [Ifound CFString but not clear if there is a one to one mapping of CFString functionality to string(3) functionality.] No one-to-one mapping, but unless you are porting code that already uses wchar_t, CFString is much more relevant. Am I missing something obvious (if so sorry!) You are probably looking for the functionality in /usr/include/wchar.h (10.3 or higher IIRC). This email sent to site_archiver(a)lists.apple.com

1 0