September 2009 - Darwin-dev

Re: Duplicate radars: good or bad? (Was: Re: Building 10.6' ld64)
by Terry Lambert 06 Sep '09

06 Sep '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Sep 4, 2009, at 10:00 AM, William Siegrist wrote: For example: On Oct 29, 2007, at 6:15 PM, Chris Espinosa wrote: On Oct 29, 2007, at 5:10 PM, David Catmull wrote: Don't let a resolution of "duplicate" keep you from filing bugs. So come to work for us. We have a jr kernel team position open. Or if you care that much, you will code. Mention me so I get the bonus. -- Terry _______________________________________________ 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… On Sep 4, 2009, at 7:03 PM, Steve Checkoway <s(a)pahtak.org> wrote: The Darwin Team is aware of the request for libunwind. I'll put down another vote for it, so no need to file more radars or emails. Comments like this confuse me since other @apple.com people have said that radars are the only way Apple knows what people are wanting and duplicate radars show more interest. On Oct 29, 2007, at 4:28 PM, Chris Hanson <cmh(a)mac.com> wrote: It ought to be just the opposite: it means you're not the only one complaining about the issue, which should be encouraging. It is, in fact. Many teams (mine included) put a premium on high- dup-count bugs, and the Mac OS X Project Office usually gives recognition and a small prize for the engineer who fixed the bug with higest dup count in a given release This email sent to site_archiver(a)lists.apple.com

1 0

Re: error -10810
by IainS 05 Sep '09

05 Sep '09

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: error -10810
by Terry Lambert 05 Sep '09

05 Sep '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com -- Terry Thanks Terry, On 5 Sep 2009, at 10:57, Terry Lambert wrote: However, to answer your specific points: ( I also confirm that I can launch new non-gui processes at will ) but all gui-based application launches fail (with that error code). thanks again, Iain _______________________________________________ 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… Oh. And top will not report zombies, which is why I specified the ps with the bsd argument list. On Sep 5, 2009, at 4:59 AM, IainS <developer(a)sandoe-acoustics.co.uk> wrote: building and checking gcc is quite stressful of system resources, it seems. Your process table is full. I am used to the temporary over-run of processes (manifesting as "can't fork"). This current situation is new to me. Any particular reason that the error numbers < -9000 are not apparent on the apple support site? should I file an enhancement request? The most likely cause is the dejagnu test suite not reaping its zombie child processes. Other possibilities are you are running the older buggy dejagnu test suite which closes the master side of the pry before the child process has drained its output, so the individual tests are stuck waiting for someone to reopen it and read their output so their closes can complete, or you have some other runaway software, or you have software that calls out to userspace from a kernel next and deadlocks itself when it makes a system call because it did not make itself immune to its own outcalls (usually firewall or antivirus software). I'm working with gcc-svn-trunk and there are some code-generation hassles right now so it's possible that the nasty is a result of those. I have the latest XCode tools, latest released dejagnu/expect and the correct gnu toolset pre-requisites for the build. There is scope for other influences on the i686 (8core Xeon) - but the powerpc (Quad G5) is a vanilla install of 10.5. ** I've repeated the effect but I can't see the bogey using either ps or activity monitor... top shows 85-ish processes, between 3 & 5 running 2 stuck and 80-ish sleeping - 320 ish threads. there is one suspicious beast (a gdb related check that hangs around) .. but killing that does not resolve the problem. one other observation is that when I log out and back in again - it is evident that the logout hasn't properly killed all subordinate processes (e.g. an ssh session left hanging.. and an iTunes instance left hanging w/out it's gui). This email sent to site_archiver(a)lists.apple.com

1 0

Re: error -10810
by Terry Lambert 05 Sep '09

05 Sep '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thanks Terry, On 5 Sep 2009, at 10:57, Terry Lambert wrote: However, to answer your specific points: ( I also confirm that I can launch new non-gui processes at will ) but all gui-based application launches fail (with that error code). thanks again, Iain _______________________________________________ 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… Not really. There is a bug in the "expect" they ship that eats a process slot per run because it doesn't actually check the test output fully. You could either say the test was useless or the expect was bogus (your choice), but it's kinds weird to run a test and not actually trap its output for analysis. -- Terry On Sep 5, 2009, at 4:59 AM, IainS <developer(a)sandoe-acoustics.co.uk> wrote: building and checking gcc is quite stressful of system resources, it seems. Your process table is full. I am used to the temporary over-run of processes (manifesting as "can't fork"). This current situation is new to me. Any particular reason that the error numbers < -9000 are not apparent on the apple support site? should I file an enhancement request? The most likely cause is the dejagnu test suite not reaping its zombie child processes. Other possibilities are you are running the older buggy dejagnu test suite which closes the master side of the pry before the child process has drained its output, so the individual tests are stuck waiting for someone to reopen it and read their output so their closes can complete, or you have some other runaway software, or you have software that calls out to userspace from a kernel next and deadlocks itself when it makes a system call because it did not make itself immune to its own outcalls (usually firewall or antivirus software). I'm working with gcc-svn-trunk and there are some code-generation hassles right now so it's possible that the nasty is a result of those. I have the latest XCode tools, latest released dejagnu/expect and the correct gnu toolset pre-requisites for the build. There is scope for other influences on the i686 (8core Xeon) - but the powerpc (Quad G5) is a vanilla install of 10.5. ** I've repeated the effect but I can't see the bogey using either ps or activity monitor... top shows 85-ish processes, between 3 & 5 running 2 stuck and 80-ish sleeping - 320 ish threads. there is one suspicious beast (a gdb related check that hangs around) .. but killing that does not resolve the problem. one other observation is that when I log out and back in again - it is evident that the logout hasn't properly killed all subordinate processes (e.g. an ssh session left hanging.. and an iTunes instance left hanging w/out it's gui). This email sent to site_archiver(a)lists.apple.com

1 0

Re: error -10810
by IainS 05 Sep '09

05 Sep '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thanks Terry, On 5 Sep 2009, at 10:57, Terry Lambert wrote: However, to answer your specific points: ( I also confirm that I can launch new non-gui processes at will ) but all gui-based application launches fail (with that error code). thanks again, Iain _______________________________________________ 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… building and checking gcc is quite stressful of system resources, it seems. Your process table is full. I am used to the temporary over-run of processes (manifesting as "can't fork"). This current situation is new to me. Any particular reason that the error numbers < -9000 are not apparent on the apple support site? should I file an enhancement request? The most likely cause is the dejagnu test suite not reaping its zombie child processes. Other possibilities are you are running the older buggy dejagnu test suite which closes the master side of the pry before the child process has drained its output, so the individual tests are stuck waiting for someone to reopen it and read their output so their closes can complete, or you have some other runaway software, or you have software that calls out to userspace from a kernel next and deadlocks itself when it makes a system call because it did not make itself immune to its own outcalls (usually firewall or antivirus software). I'm working with gcc-svn-trunk and there are some code-generation hassles right now so it's possible that the nasty is a result of those. I have the latest XCode tools, latest released dejagnu/expect and the correct gnu toolset pre-requisites for the build. There is scope for other influences on the i686 (8core Xeon) - but the powerpc (Quad G5) is a vanilla install of 10.5. ** I've repeated the effect but I can't see the bogey using either ps or activity monitor... top shows 85-ish processes, between 3 & 5 running 2 stuck and 80-ish sleeping - 320 ish threads. there is one suspicious beast (a gdb related check that hangs around) .. but killing that does not resolve the problem. one other observation is that when I log out and back in again - it is evident that the logout hasn't properly killed all subordinate processes (e.g. an ssh session left hanging.. and an iTunes instance left hanging w/out it's gui). This email sent to site_archiver(a)lists.apple.com

1 0

Re: error -10810
by Terry Lambert 05 Sep '09

05 Sep '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com You need to: (1) make sure you have two extra Terminal windows open (2) do NOT set Terminal windows to close on process exit (3) cause the problem while the extra windows are open (4) after the problem happens, issue the command: exec ps gaxlwww -- Terry Hello , on both powerpc-apple-darwin9 and i686-apple-darwin9 nothing will launch from the dock, double clicking ... or open -a. log out / back in solves the problem. cheers, Iain _______________________________________________ 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/tlambert%40apple.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… Your process table is full. The most likely cause is the dejagnu test suite not reaping its zombie child processes. Other possibilities are you are running the older buggy dejagnu test suite which closes the master side of the pry before the child process has drained its output, so the individual tests are stuck waiting for someone to reopen it and read their output so their closes can complete, or you have some other runaway software, or you have software that calls out to userspace from a kernel next and deadlocks itself when it makes a system call because it did not make itself immune to its own outcalls (usually firewall or antivirus software). In the first extra terminal window. This will replace the shell, but tell you about the processes (which you will be able to scroll back to because of #2) (5) look at who the parent process is for all the zombies. If there are no zombies, you should be able to identify a forkbomb by process name. Otherwise look for third party software to reinstall to see if you can resolve a deadlock. (6) in the second extra terminal window, shoot that process in the head (if it's a forkbomb type runaway, you can exec the kill command to use the shell process slot to do the kill in the second window; you will lose the shell but you should be able to open new windows after that, if you killed the right thing) (7) never, ever, run that process again, unless you have fixed it to reap its children correctly/not run away, and recompiled and reinstalled it On Sep 5, 2009, at 1:51 AM, IainS <developer(a)sandoe-acoustics.co.uk> wrote: I have a phenomenon I'm trying to understand. I run a large dejagnu testsuite (the gcc make check) from time to time and some amazing interaction has appeared. almost (>90%) on every occasion I run the testsuite it trashes application launch. The error message reported is [alert dialogue] "the application xxxx cannot be launched. -10810" neither google nor support on the apple website have proved illuminating on what the error message means .. and I'd like some opinions on how to debug this interaction. This email sent to tlambert(a)apple.com This email sent to site_archiver(a)lists.apple.com

1 0

Re: error -10810
by IainS 05 Sep '09

05 Sep '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On 5 Sep 2009, at 10:16, Joel Reymont wrote: Does it launch too many processes, so many that no new ones can start? On Sep 5, 2009, at 9:51 AM, IainS wrote: I just can't launch any new GUI process, it seems. cheers Iain _______________________________________________ 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… nothing will launch from the dock, double clicking ... or open -a. nope. I can run GUI-less command line processes [e.g. top in a terminal window] Already-launched GUI apps seem to continue OK (although I've seen some beach-ball behavior when trying to open a new website in Safari, this is not conclusive) BTW both systems are on 10.5.8 and latest updates (unless any updates appeared this morning) This email sent to site_archiver(a)lists.apple.com

1 0

Re: error -10810
by Joel Reymont 05 Sep '09

05 Sep '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Does it launch too many processes, so many that no new ones can start? On Sep 5, 2009, at 9:51 AM, IainS wrote: --- fastest mac firefox! http://wagerlabs.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… nothing will launch from the dock, double clicking ... or open -a. This email sent to site_archiver(a)lists.apple.com

1 0

error -10810
by IainS 05 Sep '09

05 Sep '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hello , on both powerpc-apple-darwin9 and i686-apple-darwin9 nothing will launch from the dock, double clicking ... or open -a. log out / back in solves the problem. cheers, Iain _______________________________________________ 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 phenomenon I'm trying to understand. I run a large dejagnu testsuite (the gcc make check) from time to time and some amazing interaction has appeared. almost (>90%) on every occasion I run the testsuite it trashes application launch. The error message reported is [alert dialogue] "the application xxxx cannot be launched. -10810" neither google nor support on the apple website have proved illuminating on what the error message means .. and I'd like some opinions on how to debug this interaction. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Duplicate radars: good or bad? (Was: Re: Building 10.6' ld64)
by Terry Lambert 05 Sep '09

05 Sep '09

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com The corollary to "The squeaky wheel gets the grease" is: “If trees could scream, would we be so cavalier about cutting them down? We might, if they screamed all the time, for no good reason.” - Jack Handy -- Terry _______________________________________________ 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… On Sep 5, 2009, at 12:25 AM, Joar Wingfors wrote: On 4 sep 2009, at 19.11, Terry Lambert wrote: The radar in this case isn't an engineering radar, it's a legal one, and they d not like taking interrupts. A bunch of dups on a legal radar are not going to really bear on what is, in the end, a business rather than a technical decision. And you're suggesting that business decisions aren't influenced by the amount of public interest? :-D This email sent to site_archiver(a)lists.apple.com

1 0