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

Re: Tiger speed issues...
by Justin Walker 14 May '05

14 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 14, 2005, at 16:44, Warwick Hall wrote: Hi, Or, in short form: invest in more memory :-} Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large Institute for General Semantics -------- When LuteFisk is outlawed, Only outlaws will have LuteFisk -------- _______________________________________________ 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 iBook G4 1GHz 256MB computer as my development machine. Recently I installed Tiger. Before my machine was running smoothly. Now it is noticeably slower. Waking up and application switching takes toooooo long for instance. I can feel your pain, having a 500Mhz Pismo (G3 laptop), with 256MB of memory... Is this due to the fact the Tiger kernel is 64 bit based and G4 processors are not? Or that I have 256MB RAM instead of 1GB? The kernel is not 64-bit on any version of Mac OS X/Darwin. For processors that support it (G5 only, so far), some user-mode programs are 64-bit, but the kernel is not, and may not be for a considerable while. What you are experiencing is, I think, due to the small size of memory. I think (although I've not been motivated to track this down) that there is more paging going on now, possibly because apps and frameworks are bigger. However, I have to say, once an app is up front and in control, my overworked Pismo feels way snappier than earlier versions of Mac OS X. I can churn CPU cycles like mad with long-running math computations, and still build large C++ apps in Xcode fairly quickly. The only time things bog down is when gcc decides to expand its size to double the amount of memory I have. But that's C++ for you. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Tiger speed issues...
by wadeslists＠mac.com 14 May '05

14 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… Is this due to the fact the Tiger kernel is 64 bit based and G4 processors are not? Or that I have 256MB RAM instead of 1GB? Bingo - the RAM, that is. Tiger is, as with every new incarnation of MacOS X, even more RAM hungry than the previous version. I'd recommend you upgrade to at least 640 meg - with just Mail, Safari and a few IM programs open, I usually sit at just under 512 meg used. The biggest single thing you can do to alleviate this problem is to remove the keyboard shortcut (F12) for Dashboard. Dashboard isn't launched at login - it's run the first time you invoke it. If you don't use it, it never runs, and doesn't waste 20 meg or more of memory (roughly 7 meg per widget, up to 15 for some). P.S. This is probably a wee bit off-topic for Darwin dev. Cocoa dev might be marginally more suitable, although there's plenty of forums on the web which would be much better. Wade Tregaskis (AIM/iChat, Yahoo & Skype: wadetregaskis, ICQ: 40056898, MSN, audio/video iChat & email: wadetregaskis(a)mac.com, Jabber: wadetregaskis(a)jabber.org) -- Sed quis custodiet ipsos custodes? This email sent to site_archiver(a)lists.apple.com

1 0

Tiger speed issues...
by Warwick Hall 14 May '05

14 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi, Regards, Warwick Hall _______________________________________________ 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 iBook G4 1GHz 256MB computer as my development machine. Recently I installed Tiger. Before my machine was running smoothly. Now it is noticeably slower. Waking up and application switching takes toooooo long for instance. Is this due to the fact the Tiger kernel is 64 bit based and G4 processors are not? Or that I have 256MB RAM instead of 1GB? This email sent to site_archiver(a)lists.apple.com

1 0

Re: Cross-network renaming issue in 10.4
by Finlay Dobbie 13 May '05

13 May '05

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:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=Trv0GiUmyNddYduUSZqQHY9lQjVGgtSqhbrma1KyZGeUt/2Ws4dXHsoY4JF8XS5moRCIgtuLW1CStTP/oruTzhnDIPihfGb3nrwVt4JCoGLPGyg1eoK3XSGnkOWEYtZPJHgbl0SMO3sfNCpoPq6Y0IwFx5azlqAKg4r/qNcPRxw= On 5/13/05, Justin Walker <justin(a)mac.com> wrote: > > On May 13, 2005, at 13:00, Dan Smith wrote: > > > The test application mentioned below is available for download from: > > > > http://world.std.com/~dpbsmith/CrossNetworkRenameBug.app.zip > > If this is supposed to be a bug report, you won't get much satisfaction > here (it's not an official channel to Apple). And even if it isn't, high-level applications are outside the scope of this list. > Bug reports go to <http://bugreporter.apple.com> (and also, if you > wish, see <http://www.opendarwin.org>, for their bug reporting > mechanism). General Darwin bugs do NOT belong in OpenDarwin Bugzilla. Apple never looks there, so it's not going to get fixed in any useful fashion. The only exception is if you are interested in working on resolving the issue yourself and want a place where you can track discussion, patches, integration, etc. -- Finlay _______________________________________________ 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: Cross-network renaming issue in 10.4
by Justin Walker 13 May '05

13 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 13, 2005, at 13:00, Dan Smith wrote: The test application mentioned below is available for download from: Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large Institute for General Semantics -------- Some people have a mental horizon of radius zero, and call it their point of view. -- David Hilbert -------- _______________________________________________ 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… http://world.std.com/~dpbsmith/CrossNetworkRenameBug.app.zip If this is supposed to be a bug report, you won't get much satisfaction here (it's not an official channel to Apple). Bug reports go to <http://bugreporter.apple.com> (and also, if you wish, see <http://www.opendarwin.org>, for their bug reporting mechanism). This email sent to site_archiver(a)lists.apple.com

1 0

Cross-network renaming issue in 10.4
by Dan Smith 13 May '05

13 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com The test application mentioned below is available for download from: http://world.std.com/~dpbsmith/CrossNetworkRenameBug.app.zip Entered as bug 4118699, 13-May-2005 03:49 PM Daniel P. B. Smith: Summary: Many pieces of prepress software, including ECRM's CtServer, use "hot folders" hosted on Windows systems. Client software running on the Mac copies files to the hot folder. The remote software continuously scans the hot folder for incoming files and removes them for processing. Our specific problem involves a piece of client software we must work with, which is not under our control and to which we do not have source, which copies the file into the hot folder with a .TMP extension. When the copy is complete, it generates filenames in a numeric sequence until it finds one that does not exist (e.g. TIFF04.TIF) and renames the .TMP file. We are seeing a completely reproducible problem in 10.4 8A428 which does not occur in 10.3.8 or 10.3.9. The problem occurs when, say, TIFF01.TIF, TIFF02.TIF, TIFF03.TIF, TIFF04.TIF, and TIFF05.TIF exist in the hot folder and the processing program running on the Windows box removes TIFF01.TIF for processing. The client software copies a TMP file, sees that the file TIFF01.TIF does not exist, and attempts to rename $TIFF00$.TMP to TIFF01.TIF. There is no competing activity going on in the directory except for the client software copying files to TMP and renaming them to unused names, and the remove server software removing them. Even though TIFF01.TIF truly does not exist at the time when the rename is attempted, under 10.4 the rename attempt errors out. This suggests an inconsistency between the view of the remote directory as seen by whatever layer of software performs the renaming and whatever layer of software looks to see what files exist in the directory. I have attached a REALbasic application that allows the problem to be demonstrated easily, without requiring either the real client or server software. Incidentally, the demonstration works fine on either a remote Mac using Mac OS X or a remote Windows system. The provider of the client software gives us this description of the API calls they are using: The TIFF plugin initially creates a temporary file in the target folder, with a name generated from the template "$TIFF<2unique>$.TMP", resulting in names such as: $TIFF00$.TMP, $TIFF01$.TMP etc. The TIFF plugin on Mac OS X creates the temporary file with 'FSCreateFileUnicode' - its 'whichInfo' argument does not include kFSCatInfoPermission - Only the 'textEncodingHint' and 'finderInfo' members of the 'FSCatalogInfo' structure are set. The data fork is then opened with'FSOpenFork' with a 'permissions' argument of 'fsWrPerm'. On successful completion, the file is then closed and renamed to it's final name using the RIP's Plugin-Function-Interface file rename function. The exact behaviour of this depends upon the RIP, but on OS X at present I believe it ends up using 'FSRenameUnicode'. Steps to Reproduce: ----------------- 1) Obtain two Macintoshes running Mac OS X 10.4. We'll call one the "local" system and one the "remote" system. 2) On the remote system turn on Windows Sharing and enable it for an admin account. Let's say the remote system is named "remote" and the admin account is named "d_smith." In the admin's home directory create an empty folder named "test" 3) On the local system, mount the "admin" share so that a volume named "admin" appears on the desktop. Open it in the Finder and find the "test" folder. 4) On the local system, download and expand the attached CrossNetworkRenameBug.app.zip. Launch it. A window named "Cross-network renaming bug demonstrator" opens. 5) Drag the "test" folder to the area labelled "Drag a Remote Folder Here." The path ADMIN:test appears. 6) Press the "Setup files" button. You should see output like this: reated D_SMITH:test:FILE1.txt Created D_SMITH:test:FILE2.txt Created D_SMITH:test:FILE3.txt Created D_SMITH:test:FILE4.txt Created D_SMITH:test:FILE5.txt Done. 7) Go to the remote system D_SMITH and delete the file D_SMITH:test:FILE3.txt Delete them using the Finder at the remote system system itself. Do NOT delete them using the local machine's Finder. 8) When you have done this, press the "Rename" button. Expected Results: --------------- Because the file FILE4.txt does not exist, it should be possible to rename the file. The expected program output is: Initially: D_SMITH:test:FILE2.txt exists D_SMITH:test:FILE4.txt does not exists About to try renaming After renaming attempt: D_SMITH:test:FILE2.txt does not exist D_SMITH:test:FILE4.txt exists Results are as expected. Actual Results: ------------ A message box appears saying "unexpected results." The program logs this output: Initially: D_SMITH:test:FILE2.txt exists D_SMITH:test:FILE4.txt does not exists About to try renaming error 101 renaming D_SMITH:test:FILE2.txt After renaming attempt: D_SMITH:test:FILE2.txt exists D_SMITH:test:FILE4.txt does not exists Unexpected results: Unexpected error code 101 and D_SMITH:test:FILE2.txt should not exist and D_SMITH:test:FILE4.txt should exist. _______________________________________________ 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

pthread_cond_timedwait() bug in Tiger?
by Guenter Obiltschnig 13 May '05

13 May '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com Hello, struct timeval tv; gettimeofday(&tv, NULL); struct timespec abstime; abstime.tv_sec = tv.tv_sec + milliseconds / 1000; abstime.tv_nsec = tv.tv_usec*1000 + (milliseconds % 1000)*1000000; if (abstime.tv_nsec >= 1000000000) { abstime.tv_nsec -= 1000000000; abstime.tv_sec++; } pthread_mutex_lock(&_mutex); // error handling omitted while (...) { int rc = pthread_cond_timedwait(&_cond, &_mutex, &abstime))) if (rc == ETIMEDOUT) // we never get here if milliseconds == 0 ... } Any ideas? Thanks and best regards, Günter Obiltschnig http://www.obiltschnig.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 have a strange problem with pthread_cond_timedwait() on 10.4. The following code snipped works fine on 10.3 (and a bunch of other POSIX platforms). On 10.4 however, if milliseconds == 0, the program hangs forever in pthread_cond_timedwait() instead of returning ETIMEDOUT. milliseconds == 0 basically means that the absolute time value passed to pthread_cond_timedwait() is the result of gettimeofday(). If milliseconds has a value of 10, which means that the absolute time passed to pthread_cond_timedwait() is about 10ms into the future, pthread_cond_timedwait() returns ETIMEDOUT as expected. Could it be that there is a bug in pthread_cond_timedwait() that causes it to block forever if the absolute time value passed to it lies in the past? This email sent to site_archiver(a)lists.apple.com

1 0

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 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… I am seeing the behavior you are seeing now. The cleanup handler is being called, but printf is failing (write() is returning EINTR) when called from a cleanup handler triggered by a pthread_testcancel()'d thread. I will look into this. Thank you for the example. I don't have an answer to the async cancellation case, besides "don't do that". Async cancellation is very dangerous and should be avoided. This email sent to site_archiver(a)lists.apple.com

1 0

passwordserver_sasl-59.2
by Dale Walsh 13 May '05

13 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Mta-interface: amavisd-new-2.3.1-rc3-E (2005-05-06) at daleenterprise.com -- Dale _______________________________________________ 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've recently stepped up to 10.4 server and I'm trying to expand/add a feature that has become a requirement. I've been trying to build the paswordserver_sasl source project because I need to enable the sql module but it appears that (at the very least) header files are missing as referenced during the build process. I need to keep all of the current libsasl2.2.dylib functionality to prevent crippling/breaking any current feature functionality since this library/module package is used by many apps. I've searched the source archives attempting to locate the mysterious '/AppleInternal' folder and contents yet it doesn't appear to be located anywhere. Does anyone have a solution for building this project and enabling the sql module? 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 13 mai 05, à 09:56, Matt Watson a écrit : (un-cross-posted) matt. 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… (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? Yes. compiled with -D_APPLE_C_SOURCE. I'm testing on Tiger beta A425. Could this be the issue? It actually works maybe one every 10 test.... It seems also to depends if i switch from 2 CPU to 1 CPU actually. What about the pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, NULL) issue? This email sent to site_archiver(a)lists.apple.com

1 0