June 2008 - Darwin-dev - lists.apple.com

Re: Mach-exception forwarding changed on Leopard?
by Carl Seleborg 16 Jun '08

16 Jun '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Organization: Ableton AG User-agent: Thunderbird 2.0.0.14 (Windows/20080421) Hello again, extern "C" { __attribute__ ((visibility("default"))) kern_return_t catch_exception_raise(/*...*/) { // ... } } So that was my experience of it. To sum up: 1. Visibility attribute needed to be "default". Cheers, Carl _______________________________________________ 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… Thanks to those who proposed various solutions. It turns out that the problem was a bit different. The biggest problem was that the catch_exception_raise() function and its two other versions need to have their "visibility" attribute set to "default", so the correct way to implement them was *with* the visibility attribute declaration: Not having this extra attribute declaration resulted in a linker warning in my release build (see next paragraph). The other problem I had was that my debug build is split in a bunch of dynamic libs, and these three functions were implemented in one of them. This caused whoever it is that looks these functions up not to find them. This, in turn, resulted in task_get_exception_ports() to return a NULL port (there was one returned record, but its port member was MACH_PORT_NULL). So my own exception handler was of course not called, and the application remained blocked in some default-provided implementation of these functions. The all-in-one release build did not have this problem. (Maybe this can be solved by "pre-heating" the symbols e.g. by calling those functions early on, I just thought of that but won't have time to try it) The last problem I had was with GCC not being able to break at the point where the exception occurs when my own handler is installed. I thought that it installed a crappy value for the behavior (which should be one of EXCEPTION_DEFAULT, EXCEPTION_STATE or EXCEPTION_STATE_IDENTITIY). But upon closer inspection, it turns out it was one of those values, or-ed with the constant MACH_EXCEPTION_CODES (0x80000000) which only appear in the 10.5 version of SDK. I am building on 10.5 (so GCC was built with this constant) but against the 10.4 version of the SDK which is why I did not know about that define. Now, when this bit is set, the calls to forward the exception are not exception_raise() and friends, but mach_exception_raise() and friends. It turns out however that when I just return KERN_FAILURE if this bit is set, the debugger will correctly break at the point of exception on 10.5 (not on 10.4, however!). 2. Exception handling does not play well when your catch_exception_raise() and friends functions are implemented in dynamic libraries. 3. Need to be careful when the behavior member of the already-installed exception handler has the MACH_EXCEPTION_CODES bit set. Sorry for this lengthy mail, but the documentation is so scarce that I thought this could be useful. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Diskarbitration 10.3 and 10.4
by Kevin Elliott 16 Jun '08

16 Jun '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Maybe I was misled by http://lists.apple.com/archives/Ata-scsi-dev/2005/Sep/msg00014.html - "Disk Arbitration is private on 10.3. It is public on 10.4. Disk Arbitration is otherwise the same on 10.3 and 10.4." On Jun 16, 2008, at 2:00 AM, Andy Green wrote: Ahh, I was wrong. 10.2 is when the older API was the only available API. Dan is far more authoritative on this than I am. _______________________________________________ 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: Diskarbitration 10.3 and 10.4
by Jean-Daniel Dupas 16 Jun '08

16 Jun '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Le 16 juin 08 à 11:00, Andy Green a écrit : -----Original Message----- From: Kevin Elliott [mailto:kelliott@mac.com] Sent: 16 June 2008 00:44 To: Andy Green Cc: darwin-dev(a)lists.apple.com Subject: Re: Diskarbitration 10.3 and 10.4 and the ppc version builds and runs, however there's obviously no i386 build in there. Not to damper your enthusiasm, but you need to be aware the that public API available in 10.4 is COMPLETELY different than the private API from 10.3. It is possible to build code that will work on 10.3 & 10.4, but you'd need to write your code against the 10.3 API. The older API is much more difficult to work with and a lot more arcane. Depending on what your trying to do you may be able to make it work, but if your trying to do something complex it's going to be a hard job. Unless you MUST have 10.3 support, I suggest you target 10.4 where the API is both public, and much more obvious. Which leads to the obvious question- what are you trying to do? Hi Kevin, Maybe I was misled by http://lists.apple.com/archives/Ata-scsi-dev/2005/Sep/msg00014.html - "Disk Arbitration is private on 10.3. It is public on 10.4. Disk Arbitration is otherwise the same on 10.3 and 10.4." I'm aware that the 10.2 DiskArb is completely different, so I'm not planning to support that. Cheers, Andy If you are using Xcode 2.5, you can use the old way to do it: SDKROOT_i386 = /Developer/SDKs/MacOSX10.4u.sdk SDKROOT_ppc = /Developer/SDKs/MacOSX10.3.9.sdk _______________________________________________ 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 On Jun 12, 2008, at 8:42 AM, Andy Green wrote: I can, instead, use the 10.3.9/PrivateFrameworks/DiskArb.framework, Is it possible to have a single build, which runs the public framework on 10.4 and the private one on 10.3? All I want to do is eject and unmount a disk from within a program, so I'm calling DADiskEject and DADiskUnmount (with a couple more calls to get the right objects - see http://lists.apple.com/archives/Darwin-dev/2008/May/msg00016.html) By doing my own link wrapper with dlopen/dlsym etc. it compiles and runs on 10.3, and I'll be testing that it actually works in the next couple of days. You have to use per architecture build settings. In Xcode 3.0, Open you target build settings, select the build settings you want to set (base SDK, deployment version, ...) and then click on the little gear at the bottom of the settings window, and choose "add per architecture settings" (I don't have access to the exact label, but you get the main idea). smime.p7s

1 0

RE: Diskarbitration 10.3 and 10.4
by Andy Green 16 Jun '08

16 Jun '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thread-index: AcjPQcZxWu0O226oRoSjvoDULWLs0gATPT2w Thread-topic: Diskarbitration 10.3 and 10.4 > -----Original Message----- > From: Kevin Elliott [mailto:kelliott@mac.com] > Sent: 16 June 2008 00:44 > To: Andy Green > Cc: darwin-dev(a)lists.apple.com > Subject: Re: Diskarbitration 10.3 and 10.4 > > > On Jun 12, 2008, at 8:42 AM, Andy Green wrote: > > I can, instead, use the > 10.3.9/PrivateFrameworks/DiskArb.framework, > > and > > the ppc version builds and runs, however there's obviously no i386 > > build > > in there. > > > > Is it possible to have a single build, which runs the > public framework > > on 10.4 and the private one on 10.3? > > Not to damper your enthusiasm, but you need to be aware the that > public API available in 10.4 is COMPLETELY different than the > private > API from 10.3. It is possible to build code that will work > on 10.3 & > 10.4, but you'd need to write your code against the 10.3 API. The > older API is much more difficult to work with and a lot more > arcane. > Depending on what your trying to do you may be able to make it work, > but if your trying to do something complex it's going to be a hard > job. Unless you MUST have 10.3 support, I suggest you target 10.4 > where the API is both public, and much more obvious. > > Which leads to the obvious question- what are you trying to do? > Hi Kevin, Maybe I was misled by http://lists.apple.com/archives/Ata-scsi-dev/2005/Sep/msg00014.html - "Disk Arbitration is private on 10.3. It is public on 10.4. Disk Arbitration is otherwise the same on 10.3 and 10.4." All I want to do is eject and unmount a disk from within a program, so I'm calling DADiskEject and DADiskUnmount (with a couple more calls to get the right objects - see http://lists.apple.com/archives/Darwin-dev/2008/May/msg00016.html) By doing my own link wrapper with dlopen/dlsym etc. it compiles and runs on 10.3, and I'll be testing that it actually works in the next couple of days. I'm aware that the 10.2 DiskArb is completely different, so I'm not planning to support that. Cheers, Andy _______________________________________________ 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: Diskarbitration 10.3 and 10.4
by Kevin Elliott 15 Jun '08

15 Jun '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Is it possible to have a single build, which runs the public framework on 10.4 and the private one on 10.3? Which leads to the obvious question- what are you trying to do? -Kevin Elliott _______________________________________________ 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 Jun 12, 2008, at 8:42 AM, Andy Green wrote: I can, instead, use the 10.3.9/PrivateFrameworks/DiskArb.framework, and the ppc version builds and runs, however there's obviously no i386 build in there. Not to damper your enthusiasm, but you need to be aware the that public API available in 10.4 is COMPLETELY different than the private API from 10.3. It is possible to build code that will work on 10.3 & 10.4, but you'd need to write your code against the 10.3 API. The older API is much more difficult to work with and a lot more arcane. Depending on what your trying to do you may be able to make it work, but if your trying to do something complex it's going to be a hard job. Unless you MUST have 10.3 support, I suggest you target 10.4 where the API is both public, and much more obvious. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Mach-exception forwarding changed on Leopard?
by Terry Lambert 13 Jun '08

13 Jun '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com -- Terr On Jun 13, 2008, at 2:58 AM, ward mcfarland <ward(a)megawolf.com> wrote: On 10.4, the mechanism works great. Not on 10.5, however, the application just "hangs", not processing any messages anymore (Activity Monitor says application not answering). >From the sampling, it looks like the exception-handling thread is stuck in catch_exception_raise_*() and does not want to continue. Manually turning off blocking in my exception handler did the trick. I am pretty sure you need to do the same thing. -- w _______________________________________________ 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… The crash reporter exception mechanim was changed in Leopard to reduce false reports from normally terminating processes and processes receiving normally fatal signals which have handlers that would let them recover successfully (e.g. Implemtation of memory mapping in a SIGSEGV or stack management on guard pages in a SIGBUS). You are likely looking at the wrong exception type, and still sitting on the exception for the signal path rather than the one for crashreporter. I'm not sure this is documented anywhere but the xnu source code, which is published. I implement my own exception handler and found a similar thing. The first time, it works fine. Subsequent tries hung (even force quit took 2 or 3 tries to get it to quit). A little investigation and some luck showed me that sigprocmask was turned ON for all signals except sigkill and sigstop. The opposite of what the man page says and opposite to prior behavior, when this mask was not changed. This email sent to tlambert(a)apple.com This email sent to site_archiver(a)lists.apple.com

1 0

Re: Mach-exception forwarding changed on Leopard?
by ward mcfarland 13 Jun '08

13 Jun '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On 10.4, the mechanism works great. Not on 10.5, however, the application just "hangs", not processing any messages anymore (Activity Monitor says application not answering). >From the sampling, it looks like the exception-handling thread is stuck in catch_exception_raise_*() and does not want to continue. Manually turning off blocking in my exception handler did the trick. I am pretty sure you need to do the same thing. -- w _______________________________________________ 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 implement my own exception handler and found a similar thing. The first time, it works fine. Subsequent tries hung (even force quit took 2 or 3 tries to get it to quit). A little investigation and some luck showed me that sigprocmask was turned ON for all signals except sigkill and sigstop. The opposite of what the man page says and opposite to prior behavior, when this mask was not changed. This email sent to site_archiver(a)lists.apple.com

1 0

Re: 'lookupd' fails to query DNS
by Allan Nathanson 13 Jun '08

13 Jun '08

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com - Allan On Jun 11, 2008, at 11:42 AM, Michael Grigoni wrote: Michael Grigoni wrote: Thanks for your reply; here is the output of 'scutil --dns'; I will insert a run of 'scutil' inside /etc/rc before 'lookupd' launches and post that result later: 'scutil --dns' run just prior to launching 'lookupd' (and when a ping to a FQDN works) reports; "No DNS configuration available" Output of ipconfig getoption en0 domain_name_server: ipconfig_get_option failed, (os/kern) failure Here are /Library/Preferences/SystemConfiguration/*: preferences.plist: NetworkInterfaces.plist: _______________________________________________ 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… Aside from the IP address acquired as part of the boot process, it looks as if your [netboot'd] system does not have a stored "network configuration". On Tiger systems the template configuration is part of the install image and will be updated by the Setup Assistant or by the Network pref pane. If the network has not been "configured" you won't have a DNS configuration. I'd suggest you check (and update) the /Library/Preferences/ SystemConfiguration/preferences.plist file on the netboot image. This plist should include at least one "set" (location) with at least one "service" for the Built-in Ethernet. Marc Majka wrote: Perhaps there's an error in the DNS configuration in the configuration datastore. The command "scutil --dns" will print your DNS client configuration. That would be useful to see. After startup of LaunchDaemons/* and mach_init.d/*, one or two console messages: "Workaround Bonjour, Function not implemented" is printed. FWIW, here is 'scutil; list': subKey [0] = DirectoryService:PID subKey [1] = Plugin:IPConfiguration subKey [2] = Setup: subKey [3] = Setup:/Network/HostNames subKey [4] = Setup:/System subKey [5] = State:/IOKit/PowerManagement/CurrentSettings subKey [6] = State:/Network/Interface subKey [7] = State:/Network/Interface/en0/IPv4 subKey [8] = State:/Network/Interface/en0/Link subKey [9] = State:/Network/Interface/fw0/Link subKey [10] = State:/Network/Interface/lo0/IPv4 subKey [11] = State:/Network/Interface/lo0/IPv6 subKey [12] = com.apple.DirectoryService.NotifyTypeStandard:DirectoryNodeAdded <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple Computer//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd "> <plist version="1.0"> <dict> <key>System</key> <dict> <key>Network</key> <dict> <key>HostNames</key> <dict> <key>LocalHostName</key> <string>darwin1</string> </dict> </dict> <key>System</key> <dict> <key>ComputerName</key> <string>darwin1</string> <key>ComputerNameEncoding</key> <integer>1536</integer> </dict> </dict> </dict> </plist> <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple Computer//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd "> <plist version="1.0"> <dict> <key>Interfaces</key> <array> <dict> <key>BSD Name</key> <string>en0</string> <key>IOBuiltin</key> <true/> <key>IOInterfaceType</key> <integer>6</integer> <key>IOInterfaceUnit</key> <integer>0</integer> <key>IOLocation</key> <string></string> <key>IOMACAddress</key> <data> ABEkcYZk </data> <key>IOPathMatch</key> <string>IOService:/MacRISC4PE/ht@0,f2000000/AppleMacRiscHT/pci@1/ IOPCI2PCIBridge/ethernet@F/UniNEnet/IOEthernetInterface</string> </dict> <dict> <key>BSD Name</key> <string>en1</string> <key>IOBuiltin</key> <false/> <key>IOInterfaceType</key> <integer>6</integer> <key>IOInterfaceUnit</key> <integer>1</integer> <key>IOLocation</key> <string></string> <key>IOMACAddress</key> <data> ABEkI6su </data> <key>IOPathMatch</key> <string>IOService:/MacRISC4PE/ht@0,f2000000/AppleMacRiscHT/pci@2/ IOPCI2PCIBridge/pci80211@1/AirPortPCI_MM/IOEthernetInterface</string> </dict> <dict> <key>BSD Name</key> <string>fw0</string> <key>IOBuiltin</key> <false/> <key>IOInterfaceType</key> <integer>144</integer> <key>IOInterfaceUnit</key> <integer>0</integer> <key>IOLocation</key> <string></string> <key>IOMACAddress</key> <data> ABEk//5xhmQ= </data> <key>IOPathMatch</key> <string>IOService:/MacRISC4PE/ht@0,f2000000/AppleMacRiscHT/pci@3/ IOPCI2PCIBridge/firewire@E/AppleFWOHCI/IOFireWireController/ IOFireWireLocalNode/IOFireWireIP/IOFWInterface</string> </dict> </array> </dict> </plist> Do I need to manually configure with 'scutil'? Am I wrong in assuming that 'Bonjour' should be managing the config based upon DHCP replies? During boot, the system uses the DHCP replies for nameserver, domain name, etc. until configd and lookupd start. Michael This email sent to site_archiver(a)lists.apple.com

1 0

Re: Create Process As User
by Terry Lambert 13 Jun '08

13 Jun '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com -- Terry On Jun 12, 2008, at 10:23 AM, Garth Cummings wrote: Hi Luc, On Jun 12, 2008, at 2:18 AM, Luc Andre wrote: I have a daemon (user root) that want to start an application in the current logged user environement. Best regards, --gc ____________________________________________________________________ Garth Cummings gcummings(a)apple.com Sr. Software Engineer Apple Developer Technical Support <http://developer.apple.com/technicalsupport> _______________________________________________ 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/matt.w.burnett%40gmail.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/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… KUNCExecute() is unsupported KPI as of Tiger. It exists solely for the use of old KEXTs and will likely be removed at some point in the future. On Jun 12, 2008, at 5:27 PM, Matt Burnett wrote: You could have you daemon contact your KEXT, and have the KEXT do KUNCExecute with kOpenAppAsConsoleUser. You probably don't actually want to do that, but without more details about what you're trying to accomplish I can't be sure. In any event, I highly recommend you read Technical Note TN2083 "Daemons and Agents": <http://developer.apple.com/technotes/tn2005/tn2083.html>. This email sent to matt.w.burnett(a)gmail.com This email sent to tlambert(a)apple.com This email sent to site_archiver(a)lists.apple.com

1 0

Re: Diskarbitration 10.3 and 10.4
by Terry Lambert 13 Jun '08

13 Jun '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Is it possible to have a single build, which runs the public framework on 10.4 and the private one on 10.3? -- 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 Jun 12, 2008, at 8:42 AM, Andy Green wrote: I have an app of which I'd like to deploy a single binary that will work on 10.3 and later. It links against the public DiskArbitration framework, so at the moment I can't run it on 10.3. I can, instead, use the 10.3.9/PrivateFrameworks/DiskArb.framework, and the ppc version builds and runs, however there's obviously no i386 build in there. And if so, what's the incantation? It's a dylib, so rather than linking it directly, you could provide a wrapper that opened on or the other and looked up the symbols and Did The Right Thing. You could also use install_name_tool at install time to select between them. The first option will be better if there are any header changes that make the libraries binary incompatible with each other, but is more work. If there are no header changes, you can do it in the installer by selecting which one to have it look for at install time, rather than trying to do it at runtime. This email sent to site_archiver(a)lists.apple.com

1 0