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

Re: dynamic linker problem
by Shantonu Sen 12 Jun '05

12 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Shantonu On Jun 12, 2005, at 1:38 PM, Matt Watson wrote: This email sent to ssen(a)opendarwin.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… Note that this applies only to the 32-bit linker. The 64-bit linker for ppc64 emits LC_LOAD_DYLINKER load commands when linking MH_EXECUTE outputs, and does not embed it in crt1.o how do I force the LC_LOAD_DYLINKER load command into a Mach-O executable? When crt1.o is linked, it passes /usr/lib/dyld as the first argument to the linker. You will need to do something similar if you're writing your own crt1.o substitute. matt. _______________________________________________ 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/ssen% 40opendarwin.org This email sent to site_archiver(a)lists.apple.com

1 0

Re: dynamic linker problem
by Matt Watson 12 Jun '05

12 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com matt. _______________________________________________ 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… how do I force the LC_LOAD_DYLINKER load command into a Mach-O executable? When crt1.o is linked, it passes /usr/lib/dyld as the first argument to the linker. You will need to do something similar if you're writing your own crt1.o substitute. This email sent to site_archiver(a)lists.apple.com

1 0

dynamic linker problem
by Rob Ballantyne 12 Jun '05

12 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mozilla Thunderbird 1.0 (Macintosh/20041206) Hello All, The short question is: how do I force the LC_LOAD_DYLINKER load command into a Mach-O executable? The longer version is as follows. I am writing a pure assembly routine (no C main function, no crt0.o) that I want to call dynamically linked routines (printf!). You may wonder why, of course!!! I'm just trying to understand the user land process startup with dylinker. The Mach-O Runtine Architecture document is clear about the structure of the lazy symbol stubs required to link to dynamicly loaded code. The Csu-46 package (if someone could tell me why it's called that I'd appreciate it!!!, oddest named thing I've seen) has the standard startup and runtime stuff for C (start.s, dyld.c, crt.c). I think I've got it BUT there is no LD_LOAD_DYLINKER command in the executable produced even though I've specified -dynamic everywhere. Any advice would be greatly appreciated!! 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: mistake in Tiger's python sysconfig
by Peter O'Gorman 10 Jun '05

10 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mozilla Thunderbird 1.0 (Macintosh/20041206) -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Jack Howarth wrote: | | python -c 'from distutils.sysconfig import *; print get_config_var("LINKFORSHARE | D")' | -u __dummy -u _PyMac_Error -framework System Python.framework/Versions/2.3/Pytho | n -framework CoreServices -framework Foundation It should say "-framework Python", file a bug. Peter - -- Peter O'Gorman - http://www.pogma.com -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) iQCVAwUBQqjqU7iDAg3OZTLPAQJyeAP8DAVFrmWyJN8Vv1nIP5Jikymu2XUcuofG SL0kayw8e31idZ1sSUtn0cTURV9rQCjup8V6B/b4j0BOHd1sbX9ibN3RTKWPBOOk Cotn4nv+FBYmY1+xrlrl90FFi6XMJHvUAnEKjx/yauWGCKIYUEkOn28NMhxXFU/I FR98/sJZqTs= =Cyst -----END PGP SIGNATURE----- _______________________________________________ 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

mistake in Tiger's python sysconfig
by howarth＠bromo.msbb.uc.edu 10 Jun '05

10 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com While building xplor-nih against the python framework in Tiger, I noticed what I believe to be a bug in the sysconfig results of python. In Tiger we now have a libpython2.3.dylib in /usr/lib which didn't exist before. Xplor-nih used the sysconfig call of python to find out the appropriate linker flags for linking in python. In Tiger I get... python -c 'from distutils.sysconfig import *; print get_config_var("LINKFORSHARE D")' -u __dummy -u _PyMac_Error -framework System Python.framework/Versions/2.3/Pytho n -framework CoreServices -framework Foundation ...which results in a link error when using g++ that says the file Python.framework/Versions/2.3/Python could not before found. I find that if I manually edit out Python.framework/Versions/2.3/Python from the link line and just use -lpython2.3 that g++ can link against the python framework fine. In fact, examining the linked binary with "otool -L" I see it is linked against... /System/Library/Frameworks/Python.framework/Versions/2.3/Python (compatibility v ersion 2.3.0, current version 2.3.5) ...rather than /usr/lib/libpython2.3.dylib... which actually is a tad confusing. Has anyone else noticed this issue in Tiger? Jack _______________________________________________ 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: NSInstallLinkEditErrorHandlers() does not work in Tiger.
by Leon Bottou 09 Jun '05

09 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: KMail/1.8 On Wednesday 08 June 2005 06:23 pm, you wrote: > Leon Bottou wrote: > | In Tiger, NSInstallLinkEditErrorHandlers() does not work anymore. > | CONCLUSION: > | When running on Tiger, > | the undefined handler was not called. > File a bug at <http://bugreporter.apple.com> I did that already. But I do not expect an answer (from past experience) > For a workaround, use the NSLINKMODULE_OPTION_RETURN_ON_ERROR option. This does not work when modules have cross dependencies. Suppose foo.c ------------- extern int bar(int); int foo(int x) { if (x < 1) return 1 ; else return x + bar(x-1); } bar.c ------------- extern int foo(int); int bar(int x) { if (x < 1) return 1 ; else return x + foo(x-1); } Loading either foo or bar with NSLINKMODULE_OPTION_RETURN_ON_ERROR will cause an undefined symbol error, undo the load, and return the error code. There is no suitable load order.... The problem is that I do not have the choice of restructuring the code I am loading. This is used in Lush <lush.sourceforge.net> to load foreign code into the lisp core.... The best I can do with NSLINKMODULE_OPTION_RETURN_ON_ERROR is damage control: make it work most of the time... Thanks for the suggestion anyway. - L. _______________________________________________ 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: Pending signals
by Jean-Luc Béchennec 09 Jun '05

09 Jun '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com oups, I use Mac OS X 10.3.9 :) -- Jean-Luc _______________________________________________ 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

Pending signals
by Jean-Luc Béchennec 09 Jun '05

09 Jun '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com Hi all, void tpl_get_lock(tpl_lock *lock) { /* * block the handling of signals */ if (sigprocmask(SIG_BLOCK,&signal_set,NULL) == -1) { perror("tpl_get_lock failed"); exit(-1); } } The function that unblock was the following one void tpl_release_lock(tpl_lock *lock) { /* * unblock the handling of signals */ if (sigprocmask(SIG_UNBLOCK,&signal_set,NULL) == -1) { perror("tpl_release_lock failed"); exit(-1); } } and the signal was never delivered I modified tpl_release_lock to look at pending signals : void tpl_release_lock(tpl_lock *lock) { sigset_t pending_sigs; sigemptyset(&pending_sigs); if (sigpending(&pending_sigs) < 0) { perror("tpl_release_lock failed"); exit(-1); } if (pending_sigs != 0) { printf("There is one pending\n"); } /* * unblock the handling of signals */ if (sigprocmask(SIG_UNBLOCK,&signal_set,NULL) == -1) { perror("tpl_release_lock failed"); exit(-1); } } with no success, pending_sigs is always 0 Could somebody hilight me about my problem ? Best regards -- Jean-Luc _______________________________________________ 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 working on an open source implementation of OSEK-VDX and I decided to be able to run an OSEK app in a process on my Mac (the OSEK OS run on a virtual machine which is a Unix process). I use signals to emulate hardware interrupts. Currently I am blocking signal when I enter an OSEK OS call and I unblock them when I leave. The reading of sigaction man page leaves me the impression the blocked signals are memorized and delivered when unblocked but it does not seems to work properly. The function that blocks the signals is the following one (lock is unused, signal_set is a global inited at start with the signal that are enables) : This email sent to site_archiver(a)lists.apple.com

1 0

Re: NSInstallLinkEditErrorHandlers() does not work in Tiger.
by Peter O'Gorman 08 Jun '05

08 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mozilla Thunderbird 1.0 (Macintosh/20041206) -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Leon Bottou wrote: | In Tiger, NSInstallLinkEditErrorHandlers() does not work anymore. | CONCLUSION: | When running on Tiger, | the undefined handler was not called. File a bug at <http://bugreporter.apple.com> For a workaround, use the NSLINKMODULE_OPTION_RETURN_ON_ERROR option. Peter - -- Peter O'Gorman - http://www.pogma.com -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) iQCVAwUBQqdv1biDAg3OZTLPAQLLlQQAkbTZF6FHVsjbOJIYuGS8PTwuxsApDDy+ lHbbMwL48FvVuruxDT1imbJdi8OCcruz8Sh17rZ6GlgN9g0FA52JmX6lvN57nRpe tkIjiieXAYENHZwWAsxv+1VtTG78MpL99JHDHRBzcqus0aR5d4T3+WwY0/PxxHyV hRkNPQTUuu4= =8N+k -----END PGP SIGNATURE----- _______________________________________________ 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

NSInstallLinkEditErrorHandlers() does not work in Tiger.
by Leon Bottou 08 Jun '05

08 Jun '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Resent-date: Wed, 8 Jun 2005 17:12:23 -0400 Resent-from: Leon Bottou <leonb(a)nec-labs.com> Resent-message-id: <200506081712.23634.leonb(a)nec-labs.com> Resent-to: darwin-dev(a)lists.apple.com User-agent: KMail/1.8 In Tiger, NSInstallLinkEditErrorHandlers() does not work anymore. Example: ----- begin foo.c extern int bar(int); int foo(int x) { return bar(x); } ----- end foo.c ----- begin bar.c int bar(int x) { return x*x; } ----- end bar.c ----- begin load.c #include <stdlib.h> #include <stdio.h> #include <mach-o/dyld.h> static NSLinkEditErrorHandlers handlers; static void load(const char *bundle) { NSModule module; NSObjectFileImage image; printf("Loading '%s'\n", bundle); if (NSCreateObjectFileImageFromFile(bundle,&image) == NSObjectFileImageSuccess) if (NSLinkModule(image,bundle,NSLINKMODULE_OPTION_BINDNOW)) { printf("Loaded '%s'\n", bundle); return; } printf("Problem loading %s\n", bundle); exit(10); } static void undefined(const char *sname) { printf("Entering undefined handler '%s'\n", sname); load("bar.bundle"); printf("Leaving undefined handler '%s'\n", sname); } int main(void) { handlers.undefined = undefined; NSInstallLinkEditErrorHandlers(&handlers); load("foo.bundle"); } ----- end load.c Compile with: $ cc load.c -o load $ cc -bundle -flat_namespace -undefined suppress bar.c -o bar.bundle $ cc -bundle -flat_namespace -undefined suppress foo.c -o foo.bundle UNDER PANTHER: $ ./load Loading 'foo.bundle' Entering undefined handler '_bar' Loading 'bar.bundle' Loaded 'bar.bundle' Leaving undefined handler '_bar' Loaded 'foo.bundle' UNDER TIGER: $ ./load Loading 'foo.bundle' dyld: NSLinkModule() error dyld: Symbol not found: _bar Referenced from: foo.bundle Expected in: flat namespace Trace/BPT trap CONCLUSION: When running on Tiger, the undefined handler was not called. _______________________________________________ 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