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

Re: mutex PTHREAD_MUTEX_ERRORCHECK and pthread_cancel issues
by Matt Watson 13 May '05

13 May '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com (un-cross-posted) 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… (1) using pthread_testcancel instead pthread_cond_wait in the thread loop of does not work.... Why? It works for me on Tiger. Are you sure you're building with - D_APPLE_C_SOURCE? This email sent to site_archiver(a)lists.apple.com

1 0

Re: mutex PTHREAD_MUTEX_ERRORCHECK and pthread_cancel issues
by Stéphane Letz 13 May '05

13 May '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com Le 12 mai 05, à 22:27, Matt Watson a écrit : On May 12, 2005, at 11:25 AM, Stéphane Letz wrote: Are you sure your code has reached a cancellation point? The following works fine for me on Tiger: % cat test.c #include <assert.h> #include <pthread.h> #include <stdio.h> static pthread_mutex_t m = PTHREAD_MUTEX_INITIALIZER; static pthread_cond_t c = PTHREAD_COND_INITIALIZER; static pthread_key_t k; void * func(void *arg) { assert(!pthread_setspecific(k, arg)); assert(!pthread_mutex_lock(&m)); while (1) { assert(!pthread_cond_wait(&c, &m)); } /* NOT REACHED */ assert(!pthread_mutex_unlock(&m)); return NULL; } void cleanup_func(void *arg) { printf("cleanup called with %p\n", arg); } int main(void) { pthread_t t; int i = 42; void *ret; assert(!pthread_key_create(&k, cleanup_func)); printf("creating a thread with arg %p\n", &i); assert(!pthread_create(&t, NULL, func, &i)); assert(!pthread_cancel(t)); assert(!pthread_join(t, &ret)); assert(PTHREAD_CANCELED == ret); return 0; } % make test CFLAGS='-Os -Wall -W -D_APPLE_C_SOURCE' cc -Os -Wall -W -D_APPLE_C_SOURCE test.c -o test % ./test creating a thread with arg 0xbffffc7c cleanup called with 0xbffffc7c Thanks. It works OK. But : #include <assert.h> #include <pthread.h> #include <stdio.h> static pthread_mutex_t m = PTHREAD_MUTEX_INITIALIZER; static pthread_cond_t c = PTHREAD_COND_INITIALIZER; static pthread_key_t k; void * func(void *arg) { assert(!pthread_setspecific(k, arg)); assert(!pthread_mutex_lock(&m)); while (1) { pthread_testcancel(); } /* NOT REACHED */ assert(!pthread_mutex_unlock(&m)); return NULL; } void cleanup_func(void *arg) { printf("cleanup called with %p\n", arg); } int main(void) { pthread_t t; int i = 42; void *ret; assert(!pthread_key_create(&k, cleanup_func)); printf("creating a thread with arg %p\n", &i); assert(!pthread_create(&t, NULL, func, &i)); sleep(1); printf("pthread_cancel 0\n"); assert(!pthread_cancel(t)); printf("pthread_cancel 1\n"); printf("pthread_join 0\n"); assert(!pthread_join(t, &ret)); printf("pthread_join 1 %x\n",ret); assert(PTHREAD_CANCELED == ret); return 0; } ./test creating a thread with arg 0xbffffc20 pthread_cancel 0 pthread_cancel 1 pthread_join 0 .... (1) Using an asynchronous cancellation mode does not work also : #include <assert.h> #include <pthread.h> #include <stdio.h> static pthread_mutex_t m = PTHREAD_MUTEX_INITIALIZER; static pthread_cond_t c = PTHREAD_COND_INITIALIZER; static pthread_key_t k; void * func(void *arg) { assert(!pthread_setspecific(k, arg)); assert(!pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, NULL)); assert(!pthread_mutex_lock(&m)); while (1) { pthread_testcancel(); } /* NOT REACHED */ assert(!pthread_mutex_unlock(&m)); return NULL; } void cleanup_func(void *arg) { printf("cleanup called with %p\n", arg); } int main(void) { pthread_t t; int i = 42; void *ret; assert(!pthread_key_create(&k, cleanup_func)); printf("creating a thread with arg %p\n", &i); assert(!pthread_create(&t, NULL, func, &i)); sleep(1); printf("pthread_cancel 0\n"); assert(!pthread_cancel(t)); printf("pthread_cancel 1\n"); printf("pthread_join 0\n"); assert(!pthread_join(t, &ret)); printf("pthread_join 1 %x\n",ret); assert(PTHREAD_CANCELED == ret); return 0; } ./test creating a thread with arg 0xbffffc20 pthread_cancel 0 pthread_cancel 1 pthread_join 0 .... Thanks Stephane Letz _______________________________________________ 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… But after a pthread_cancel() has been called, the __pthread_destructor code is never called. Is there any known issue in using the pthread_key_create system? (1) using pthread_testcancel instead pthread_cond_wait in the thread loop of does not work.... Why? Is pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, NULL) supposed to work on Tiger? This email sent to site_archiver(a)lists.apple.com

1 0

Re: Problem with dlsym
by Alexey Proskuryakov 13 May '05

13 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Thu, 12 May 2005 14:49:25 -0700 "Jordan K. Hubbard" <jkh(a)apple.com> wrote: - 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… And the answer is that this is a known bug ("rdar://problem/4057081 infinte loop in dlsym(RTLD_NEXT, ) when symbol not found.", JFYI for those with radar access). Could it also be a reason for iODBC to malfunction in 10.4 (rdar://4069775)? The stack trace shows a lot of recursion in dlsym... According to the developer, it only happens when there are circular dependencies with libraries. The work around is 1) remove the circular dependency, or 2) use dlsym(RTLD_DEFAULT, "xxx") instead of RTLD_NEXT. Unfortunately, this workaround doesn't apply if one intends to use Apple's libiodbc (though, one can always revert to OpenLink's version). This email sent to site_archiver(a)lists.apple.com

1 0

Re: Problem with dlsym
by Peter O'Gorman 13 May '05

13 May '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 Peter O'Gorman wrote: | | As an alternative to Jordan's suggestion, change calls to dlsym to: | if (NSIsSymbolNameDefined(sym)) dlsym(handle,symbol); It is amazing how I could forget the underscore here and remember it below. Above won't work. | void* my_dlsym(void* module,const char* symbol) { | ~ char* u_sym=NULL; | ~ void* address=NULL; | ~ int retCode= asprintf(&u_sym,"_%s",symbol); // ignoring error, naughty! Peter -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) iQCVAwUBQoPyp7iDAg3OZTLPAQKNPwQAnl3ebS3zQ3VFL7Ruub0X2RQOJikOqqtg 4vcFnolTCfyHQ+2vzGfpxfcKYpthZMA6sQ4LfDJaiu5hMtmsNEEuK8hHTLb6W3kl kn7Tys+/V9/2Ox1thnvFEGHANxv+0jGg30jEzzvd5SKQADDaxKONJeOmzZhBfW9L 3Km385JGb6Y= =1oeA -----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

Re: Problem with dlsym
by Peter O'Gorman 12 May '05

12 May '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 J. Davison de St. Germain wrote: | We have been using the 'dlcompat' library on OSX for a while and it | worked great. However, with Tiger 10.4, it appears that the dl | functions (eg: dlsym) are now built in. Unfortunately, when we try to | find a non-existent symbol in one of our large shared libraries, it | now takes forever (instead of returning null fairly | quickly/immediately). After about 20 minutes of 'working' on the | dlsym call, our program will return and say that it couldn't find the | symbol. When we used dlcompat before, we didn't have this problem. | We have read the man page and tried all the different options that are | presented, but nothing seems to help. As an alternative to Jordan's suggestion, change calls to dlsym to: if (NSIsSymbolNameDefined(sym)) dlsym(handle,symbol); Another alternative, very hacky, works because of an implementation detail that may change, change calls to dlsym to use NSLookupSymbolInModule(): void* my_dlsym(void* module,const char* symbol) { ~ char* u_sym=NULL; ~ void* address=NULL; ~ int retCode= asprintf(&u_sym,"_%s",symbol); // ignoring error, naughty! ~ address=NSAddressOfSymbol(NSLookupSymbolInModule(handle,u_sym)); ~ free(u_sym); ~ return address; } | | (Note, if we try the same lookup of a non-existent symbol on a | system library, it returns very quickly. The library that we are | searching is part of a large framework of other libraries and (apparently) | for some reason loads all of the other libraries in its search for the | symbol. This does seem strange as we specify the exact dylib that we | want to searh through.) This is correct behavior. Well, correct except for the fact that it takes forever :). The specification for dlsym states that it will look for the symbol in objects loaded as a result of loading the object specified by handle. dlcompat also does this, but only checks one level deep (bad!) so does not get into a loop for circular dylibs. | | In a last ditch effort, we tried compiling and using the original | 'dlcompat' libraries. Yeah, this is unlikely to work. I am not convinced that the bug requires circularity. Do you have any circular dependencies in your framework? This was also noticed by fink's kde maintainer, kde startup times went > 24 hours with tiger, but I did not see any obvious circularity in the libraries. Haven't had a chance to look closely. Perhaps your setup is easier to make a smallish test case? Peter - -- Peter O'Gorman - http://www.pogma.com -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) iQCVAwUBQoPkxbiDAg3OZTLPAQJx2AP6Aqfu6C9wfkvnxNMZ1PPhjWssmD4KonnD jDLRw3YnbEYaGDlA1X7Llyna8OMat+RpQewo5MAsLi/wX2xRmeh2kJgDIypsMtJb v3+zmjizJQVJm7m+0rARApF50PkEXRIJrPz2WysSckAcVOEflZGL+JtsWZflCEZ8 vYWFI+bDCLA= =nFSB -----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

Re: Patch to mount_nfs to make NFSMNT_RESVPORT default
by Craig Rodrigues 12 May '05

12 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mutt/1.5.9i Hi, OK, I submitted the following patch to Apple, Radar # 4117658. I hope it goes in, because it improves interoperability. -- Craig Rodrigues rodrigc(a)crodrigues.org Index: src/diskdev_cmds/mount_nfs.tproj/mount_nfs.8 =================================================================== RCS file: /Volumes/src/cvs/od/src/diskdev_cmds/mount_nfs.tproj/mount_nfs.8,v retrieving revision 1.1.1.3 diff -u -r1.1.1.3 mount_nfs.8 --- src/diskdev_cmds/mount_nfs.tproj/mount_nfs.8 13 Dec 2003 15:34:40 -0000 1.1.1.3 +++ src/diskdev_cmds/mount_nfs.tproj/mount_nfs.8 12 May 2005 14:42:40 -0000 @@ -102,11 +102,15 @@ Any attempt to perform file locking operations on this mount will return the error EOPNOTSUPP regardless of whether or not the NFS server supports NFS file locking. +.It Fl N +Do +.Em not +use a reserved socket port number (see below). .It Fl P Use a reserved socket port number. -This is useful for mounting servers that require clients to use a -reserved port number on the mistaken belief that this makes NFS -more secure. (For the rare case where the client has a trusted root account +This flag is obsolete, and only retained for compatibility reasons. +Reserved port numbers are used by default now. +(For the rare case where the client has a trusted root account but untrustworthy users and the network cables are in secure areas this does help, but for normal desktop clients this does not apply.) .It Fl R Index: src/diskdev_cmds/mount_nfs.tproj/mount_nfs.c =================================================================== RCS file: /Volumes/src/cvs/od/src/diskdev_cmds/mount_nfs.tproj/mount_nfs.c,v retrieving revision 1.1.1.2 diff -u -r1.1.1.2 mount_nfs.c --- src/diskdev_cmds/mount_nfs.tproj/mount_nfs.c 13 Dec 2003 15:34:40 -0000 1.1.1.2 +++ src/diskdev_cmds/mount_nfs.tproj/mount_nfs.c 12 May 2005 14:42:41 -0000 @@ -149,7 +149,7 @@ 0, (u_char *)0, 0, - NFSMNT_NFSV3, + NFSMNT_NFSV3 | NFSMNT_RESVPORT, NFS_WSIZE, NFS_RSIZE, NFS_READDIRSIZE, @@ -305,6 +305,9 @@ case 'l': nfsargsp->flags |= NFSMNT_RDIRPLUS; break; + case 'N': + nfsargsp->flags &= ~NFSMNT_RESVPORT; + break; #ifdef NFSKERB case 'm': (void)strncpy(realm, optarg, REALM_SZ - 1); @@ -361,7 +364,7 @@ altflags = 0; break; case 'P': - nfsargsp->flags |= NFSMNT_RESVPORT; + /* obsolete for NFSMNT_RESVPORT, now default */ break; #ifdef ISO case 'p': _______________________________________________ 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: Problem with dlsym
by Jordan K. Hubbard 12 May '05

12 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.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: Patch to mount_nfs to make NFSMNT_RESVPORT default
by Andrew Gallatin 12 May '05

12 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Craig Rodrigues writes: > On Thu, May 12, 2005 at 04:04:46PM -0400, Andrew Gallatin wrote: > > AFAIK, the problem is really Apple / Linux. FreeBSD and Solaris don't > > Nope, I encountered this between MacOS X 10.3.9 (NFS client) > and FreeBSD 4.9 (NFS server) and SuSE Enterprise Linux 9 (NFS server). > > i.e. on MacOS X, I had to do a mount_nfs -P in order to > mount exports on both of those servers. Strange. 5.x has vfs.nfsrv.nfs_privport which defaults to off, and I don't have problems with mounts from a FreeBSD 5 server. FreeBSD 4.x has vfs.nfs.nfs_privport which I *thought* was the same thing. And it it also defaults to off.. 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: Setting file metadata, how?
by Vince DeMarco 12 May '05

12 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 11, 2005, at 6:08 PM, robert delius royar wrote: This email sent to demarco(a)apple.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… I would like to add the ability to set kMDItemKind and kMDItemContentType from an X11 application (XEmacs). I have written an mdimporter that lets spotlight index elisp files and other files I have registered though the importer (types el, mss, and text <key>public.filename-extension</key>). I created a <key>org.xemacs.ostype</key> <string>XEma</string> and a <key>UTTypeDescription</key> <string>XEmacs Document</string> <key>UTTypeIdentifier</key> <string>org.xemacs.text</string> The importer works fine except that XEmacs does not have any code in it (yet) to make sure its files have the correct metadata when they are written. What is the minimum I would need to add the ability to accomplish this? Could the ability be created in a plugin that provided some external hook that XEmacs could call with a complete path? I am thinking maybe the sample metadate application could be adapted for that. I think that would be better than having to create an Xcode wrapper for XEmacs. The minimum is to declare the UTI type for elisp files in XEmacs along with the Kind String (See the Launch Services documentation) you don't need to write an importer to do what you are asking. Ultimately, in addition to automatic indexing, I would like XEmacs to be able to determine modes for files from the metadata which would be helpful for a number of reasons. -- Dr. Robert Delius Royar Associate Professor of English Morehead State University Morehead, Kentucky _______________________________________________ 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/demarco%40apple.com This email sent to site_archiver(a)lists.apple.com

1 0

Re: Patch to mount_nfs to make NFSMNT_RESVPORT default
by Craig Rodrigues 12 May '05

12 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mutt/1.5.9i On Thu, May 12, 2005 at 07:44:07PM +0200, Markus Hitter wrote: > quickly. Making your patch compatible to darwinbuild would help a lot, > probably. I will follow the steps that you mentioned. My patch is compatible with darwinbuild. I put my patch inside a file: /usr/local/darwinbuild/Sources/diskdev_cmds-277.4-patches.tar.gz and then I built the diskdev_cmds package using the darwinbuild script (no mean feat!). Since my patch comes from the FreeBSD mount_nfs, I will submit it to Apple via Radar, since it solved a legitimate interoperability problem for me. My automount patch is a separate issue. -- Craig Rodrigues rodrigc(a)crodrigues.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