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

Re: Tiger and libreadline
by Boyd Waters 06 May '05

06 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 6, 2005, at 1:07 AM, Martin Costabel wrote: I maintain that this is broken. Are there problems with installing GNU readline? ~ boyd Boyd Waters National Radio Astronomy Observatory Socorro, New Mexico _______________________________________________ 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… Many GNU packages rely on libreadline. It seems that in Darwin there is a BSD libedit, which is symlinked as libreadline and has a corresponding readline.h. This breaks Octave and probably other builds. ... In addition, there is no BSD version of libreadline. BSD libedit has been in Mac OS X since at least 10.2. The new things are the compatibility wrappers in readline/readline.h and, and this seems to me to be Apple's invention, the symlinks readline/history.h- >readline.h and libreadline.dylib->libedit.2.dylib. The overall effect is the same (licensing issues aside) as having included a very old incompatible version of GNU readline. I used the Gentoo package manager to install readline on my Tiger system; this replaced /usr/include/readline/* with GNU's version, and replaced the symlink to libedit.dylib with a "real" version of libreadline. I did not notice any breakage of other things, but have not compiled any Cocoa apps. The only things that I have built with xCode, since making this change, have been device drivers - nothing that relied upon readline. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Darwin 8.0 Source Code Available
by howarth＠bromo.msbb.uc.edu 06 May '05

06 May '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com Kevin, You answer brings up the following question. When Panther shipped, gcc-3.3 had only had a short 10.2 based release. So I assume the original MacOS X 10.3 release was built with gcc-3.1 despite the presence of the updated gcc-3.3 compiler as default. Does anyone know how much of Panther was eventually rebuilt with gcc-3.3 by the time of the final update (10.3.9)? Did they ever build the kernel with gcc-3.3 and how many updates later was that transition? 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: Darwin 8.0 Source Code Available
by Kevin Van Vechten 06 May '05

06 May '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com On May 6, 2005, at 12:11 PM, Kevin Van Vechten wrote: - Kevin _______________________________________________ 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… The simple answer is that nothing on Mac OS X 10.4 was built with gcc-4.0. I should amend this to say that nothing available as open source on Mac OS X 10.4 was built with gcc-4.0. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Darwin 8.0 Source Code Available
by howarth＠bromo.msbb.uc.edu 06 May '05

06 May '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com Shantonu, You might notice it as a speed increase though. In the tests I have run at -O3 gcc-4.0 often creates code that can be 20-30% faster. However I have no idea how much of the kernel is already assembly. 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: Darwin 8.0 Source Code Available
by Kevin Van Vechten 06 May '05

06 May '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com - Kevin On May 6, 2005, at 8:55 AM, Jack Howarth 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… There probably are ways to tell, but I have to wonder why that's important. The simple answer is that nothing on Mac OS X 10.4 was built with gcc-4.0. Historically, major upgrades to the OS gcc compiler (i.e. the compiler used to compile the OS, gcc_os) have only occurred between major releases. There are always binary compatibility concerns when changing compilers. "If it ain't broke, don't fix it" is the software update mantra. Of course, past performance is no indication of future results... Kevin, Thanks for the clarification. I still wonder if there is some symbol in a binary or library that would indicate if it is built with gcc 3.3 or gcc 4.0. It does make one wonder if there will be a migration, through system updates, during the lifespan of Tiger where more components are built with gcc-4.0. Or is it likely such a transition to gcc-4.0 builds will only occur at MacOS X 10.5? Jack This email sent to site_archiver(a)lists.apple.com

1 0

Re: Darwin 8.0 Source Code Available
by Shantonu Sen 06 May '05

06 May '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com Shantonu On May 6, 2005, at 8:55 AM, Jack Howarth 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… IOKit is binary compatible between gcc 3.3-built and 4.0-built kernels and kexts. If there is a transition, you shouldn't notice it. Kevin, Thanks for the clarification. I still wonder if there is some symbol in a binary or library that would indicate if it is built with gcc 3.3 or gcc 4.0. It does make one wonder if there will be a migration, through system updates, during the lifespan of Tiger where more components are built with gcc-4.0. Or is it likely such a transition to gcc-4.0 builds will only occur at MacOS X 10.5? 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/ssen% 40opendarwin.org This email sent to site_archiver(a)lists.apple.com

1 0

Re: SCDynamicStore ConsoleUser change callback not working on Tiger?
by Stéphane Sudre 06 May '05

06 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com vendredi 6 mai 2005, à 10:05 AM, Daniel Becker a écrit : Hi all, I would check: _______________________________________________ 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 have a daemon which registers a callback with SCDynamicStore to get notified of console user changes. This seems to work just fine on Panther, but it appears the callback never gets triggered on Tiger. Were there any relevant changes in Tiger's SCDynamicStore that I should be aware of? Any other suggestions? (The relevant code is given below.) - that the daemon is really launched on Tiger (if it's launched by a StartupItem, due to permission enforcement changes, this might be an explanatio). - that the daemon is launched after CoreServices is enabled. This was a "funny" discovery with Tiger, you can have a StartupItem be launched before CoreServices is enabled and in this case, you can say bye bye to your daemon if it calls some CoreFoundation APIs (for instance related to CFMessagePort). This email sent to site_archiver(a)lists.apple.com

1 0

Re: Darwin 8.0 Source Code Available
by howarth＠bromo.msbb.uc.edu 06 May '05

06 May '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com Kevin, Thanks for the clarification. I still wonder if there is some symbol in a binary or library that would indicate if it is built with gcc 3.3 or gcc 4.0. It does make one wonder if there will be a migration, through system updates, during the lifespan of Tiger where more components are built with gcc-4.0. Or is it likely such a transition to gcc-4.0 builds will only occur at MacOS X 10.5? 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

SCDynamicStore ConsoleUser change callback not working on Tiger?
by Daniel Becker 06 May '05

06 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi all, Daniel void ConsoleUserChangedCallBackFunction(SCDynamicStoreRef store, CFArrayRef changedKeys, void * context) { [...] } _______________________________________________ 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 have a daemon which registers a callback with SCDynamicStore to get notified of console user changes. This seems to work just fine on Panther, but it appears the callback never gets triggered on Tiger. Were there any relevant changes in Tiger's SCDynamicStore that I should be aware of? Any other suggestions? (The relevant code is given below.) int main (int argc, const char * argv[]) { SCDynamicStoreContext context = {0, NULL, NULL, NULL, NULL}; SCDynamicStoreRef session = NULL; CFStringRef consoleUserNameChangeKey = NULL; CFArrayRef notificationKeys; CFRunLoopSourceRef rls = NULL; FILE * pid_file; daemon(0,0); syslog(1, "Daemon starting up with PID %d.\n", getpid()); session = SCDynamicStoreCreate(NULL, CFSTR(kProgramName), ConsoleUserChangedCallBackFunction, &context); if (session == NULL) { syslog(1, "Couldn't create DynamicStore session!\n"); return 1; } consoleUserNameChangeKey = SCDynamicStoreKeyCreateConsoleUser(NULL); if (consoleUserNameChangeKey == NULL) { CFRelease(session); syslog(1, "Couldn't create ConsoleUser key!\n"); return 1; } notificationKeys = CFArrayCreate(NULL, (void*)&consoleUserNameChangeKey, (CFIndex)1, &kCFTypeArrayCallBacks); CFRelease(consoleUserNameChangeKey); if (notificationKeys == NULL) { CFRelease(session); syslog(1, "Couldn't create notification key array!\n"); return 1; } if (SCDynamicStoreSetNotificationKeys(session, notificationKeys, NULL) == FALSE) { CFRelease(notificationKeys); CFRelease(session); syslog(1, "Couldn't set up dynamic store notification!\n"); return 1; } CFRelease(notificationKeys); rls = SCDynamicStoreCreateRunLoopSource(NULL, session, (CFIndex)0); pid_file = fopen(kPIDFile, "w"); fprintf(pid_file, "%d\n", getpid()); fclose(pid_file); CFRunLoopAddSource(CFRunLoopGetCurrent(), rls, kCFRunLoopDefaultMode); CFRelease(rls); CFRunLoopRun(); unlink(kPIDFile); return 0; } This email sent to site_archiver(a)lists.apple.com

1 0

Re: Tiger and libreadline
by Martin Costabel 06 May '05

06 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mozilla Thunderbird 1.0 (Macintosh/20041206) On 01/05/2005, at 6:56 PM, Martin Costabel wrote: It is, however, incompatible. There is no `BSD version' versus `GNU version'. The BSD libedit resembles readline in that they both provide emacs and vi editing modes, but there is neither source nor object compatibility. There has been some attempt to provide both, but it's not there yet (try key binding or multibyte character support). I maintain that this is broken. -- Martin _______________________________________________ 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… chet(a)cwru.edu wrote: Many GNU packages rely on libreadline. It seems that in Darwin there is a BSD libedit, which is symlinked as libreadline and has a corresponding readline.h. This breaks Octave and probably other builds. Yes, this is annoying, but it fits into the pattern of Apple first putting a broken version of a new open source package into OSX (there are quite a few historical examples for this). A correct version can probably be expected for the OSX release "Sabre-tooth cat" :-) The Tiger version is neither broken nor 'incorrect'. It does what it says on the tin. It is however the BSD version not the GNU version. In addition, there is no BSD version of libreadline. BSD libedit has been in Mac OS X since at least 10.2. The new things are the compatibility wrappers in readline/readline.h and, and this seems to me to be Apple's invention, the symlinks readline/history.h->readline.h and libreadline.dylib->libedit.2.dylib. The overall effect is the same (licensing issues aside) as having included a very old incompatible version of GNU readline. This email sent to site_archiver(a)lists.apple.com

1 0