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

Re: __builtin_frame_address()
by Steve Checkoway 28 May '05

28 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 25, 2005, at 7:57 PM, Steve Checkoway 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… This email sent to site_archiver(a)lists.apple.com Is __builtin_frame_address() broken? I have some backtracing code that walks the stack essentially using: By the lack of response here, I suspect that this might be the wrong list to ask. Can someone suggest a better one? On a slightly related note, the first few values that frame->lr holds are in unrelated functions that are not on the stack at the time. This is new in gcc 4. Prior versions seemed to provide sane addresses. After the first few functions, the stack seems correct. As a follow up to this, the addresses are correct, it was my address to symbol lookup that for some reason isn't working any more. - Steve smime.p7s

1 0

Re: can virtual memory be suppressed?
by Dave Zarzycki 28 May '05

28 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com davez On May 28, 2005, at 6:48 AM, Roland Silver wrote: This email sent to zarzycki(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… Not really. Virtual memory is fundamental to the operating system. Try running the fs_usage tool as root to figure out what is thrashing your disk with so much I/O. I have an obscure bug in a program running under 10.4, which I suspect may be caused by thrashing, even though I am running on a PowerBook with 2 GB or RAM. Is there any way I can force the operating system to let my program run with virtual memory/ memory mapping suppressed? _______________________________________________ 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/zarzycki%40apple.com This email sent to site_archiver(a)lists.apple.com

1 0

can virtual memory be suppressed?
by Roland Silver 28 May '05

28 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com I have an obscure bug in a program running under 10.4, which I suspect may be caused by thrashing, even though I am running on a PowerBook with 2 GB or RAM. Is there any way I can force the operating system to let my program run with virtual memory/ memory mapping suppressed? _______________________________________________ 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: IOHIDFamily compilation under 10.4.1
by Nicolas Diez 28 May '05

28 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:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=R0h8QLsKOkL4dOJ/TFk+nB6nJUclH0sJ5e2TnkK34AM9ikTFBwLr9Z8rK4pvBuRf6tJFYCWIf1OcAQANSMNfnGfRPCxXEsev66DRInAOdR5NTATG4RTwDYXuzQiOe7xmJonGNCBnB63TTOdNmd5/O3KNOMHfTb1JbACjSayfWHw= Well, I post what I have find : In fact, gcc4.0 is too restrictive and it appears that darwin source is compiled with gcc3.3 by Apple. Before I knew that, I have made explicit casts and redefinition of the functions and it's work. The other, more simple solution is to choose gcc3.3 as version of gcc for this project. _______________________________________________ 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: LDAP, DirectoryService and loginwindow in Tiger
by Michael Bartosh 28 May '05

28 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 20, 2005, at 3:12 PM, Finlay Dobbie wrote: The manifestation of this is that you can't log in to the GUI whilst authenticating against an LDAP server whose user accounts have passwords stored in SHA-1 hashes, but you can log in using ssh (and possibly other stuff). -mb _______________________________________________ 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… Don't map the password attribute in the LDAPv3 Plug-in. DirectoryService will perform an LDAP bind to authent the user. iirc it will to aCRAM-MD5 bind if the server is capable. If I were you I'd disable clear-text binds. Or use ssl. At any rate giving the OS access to the hashes is a bad idea, since someone could brute force them. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Bizarre out of memory bug
by Justin Walker 27 May '05

27 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Truly weird behavior... On May 27, 2005, at 1:26, Tim Seufert wrote: Cheers, Justin -- Justin C. Walker, Curmudgeon-At-Large Institute for General Semantics -------- It's not whether you win or lose... It's whether *I* win or lose. -------- _______________________________________________ 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… Things found through further experimentation: [snip] 4. The bug goes away if I don't invoke Wilfredo Sanchez's example tcsh setup (/usr/share/tcsh/examples/rc) in my .cshrc . Before you hit gdb, have you tried isolating what in Fred's setup might be causing the problem (e.g., 'echo Here I am >> log-file')? That may narrow it down enough to ring bells (I know something like this has happened with shell startup scripts, in a previous lifetime). My archives don't 5. I still only get this when SSHing from the PowerBook to the G5. Despite trying to configure the PowerBook with an identical login name and UID/GID, identical network config, and identical .cshrc, everything works fine when logging in to the PowerBook. Unless I figure out what's unique about the G5's configuration, this may not be easy to replicate reliably on other systems. Could be a "good garbage vs. bad garbage" issue (on one system, you get zeros in memory; on another, non-zero bits. If those bits are a pointer, in one case "boom"; in the other, harmless junk). I think my next step is to try to build tcsh with debugging symbols, attach to it with GDB when I log in, and step through the code to see exactly where things go wrong. I may also try to track down exactly which part of the example rc causes the problem. (Presumably it's the section which sets up completion.) If you've never debugged one of these shells before, you might consider an alternative: poke yourself in the eye with a sharp stick; it takes less time, but gives you the same pain :-} Andrew's suggestion to use ktrace is a good one. There was a similar thread back in 2003 (Guiliano Gavazzi: "tcsh: Out of Memory") that involved tcsh closing down an 'ssh' session (darwin-development list), and he seemed to have nailed down the problem with ktrace. This email sent to site_archiver(a)lists.apple.com

1 0

Re: UDF with resource fork, help needed
by Alex v.Below 27 May '05

27 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Bye Alex Am 27.05.2005 um 10:15 schrieb mlists(a)ugsoft.de: Hi, The problem is including resource forks. Has anybody out there any experience with this? Thanks for any help and ideas. Uwe _______________________________________________ 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 know a lot of stuff about UDF, but it's a been awhile. I will contact you off-list, and look into the issue I am trying to adjust mkisofs to create UDF images containing Finder file infos and resource forks, etc. Including Finder infos is already done. Converting decomposite unicode HFS+ filenames (special characters) to corresponding UDF, Joliet, Rockridge and HFS filename representation is also done. >From the ECMA 167 and UDF260 specs documentation I cannot understand how the required Extended File Entries and corresponding Stream Directory shall be implemented. E.g., does every UDF directory contain a Stream (sub)Directory with the Extended File Entries, or is the Stream Directory seperated from the UDF dir tree, or what, etc.? All I tried with resource forks resulted in system crashes (You need to restart...) when mounting the UDF image on 10.3.9 ;-( which really makes developing hard. I've also tried to find sample code how to read an UDF CD/image to find out where all the resource fork related pieces live and how they are linked together. E.g. I tried to find the Darwin source for mount_udf with no success. Any idea where I can find this or something similar showing how the resource fork stuff is accessed/created? _______________________________________________ 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/ newsletters%40vonbelow.com This email sent to newsletters(a)vonbelow.com This email sent to site_archiver(a)lists.apple.com

1 0

Re: Bizarre out of memory bug
by Tim Seufert 27 May '05

27 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Things found through further experimentation: [yobgorgle:~] tas% l /Sys<tab> I get the expected expansion to: [yobgorgle:~] tas% l /System/ So, arguments to commands don't cause the bug. _______________________________________________ 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. Bonjour is a red herring. Alexey Proskuryakov suggested forcing ssh to use IPv4 since ssh'ing to a Bonjour address will default to an IPv6 connection. And indeed I find that the problem goes away with "ssh -4 yobgorgle.local". Furthermore, if I ssh directly to Yobgorgle's IPv6 address, the problem is present. So I think Bonjour has been eliminated from culpability; it's really a question of whether the SSH session uses IPv4 or IPv6. 2. To test bash I created a fresh user account on the problem machine and did not alter its login shell setting. The result: bash works fine inside an IPv6 ssh session. So this seems to be a tcsh problem. 3. Along those lines, further observation of tcsh's behavior strongly suggests a bug in its command completion code. Pressing tab with no characters typed is OK, but with one or more characters typed tcsh will die immediately. If completion is not used, the only commands which do not crash tcsh with an 'Out of memory' error are those whose names are only 1 character, such as 'w'. Single- character aliases work too, even if they expand to multiple characters. Note, however, that all this applies only to the command name itself. If I do something like this: If I then press return, I get what I expect given that 'l' is an alias to 'ls -l'. 4. The bug goes away if I don't invoke Wilfredo Sanchez's example tcsh setup (/usr/share/tcsh/examples/rc) in my .cshrc . 5. I still only get this when SSHing from the PowerBook to the G5. Despite trying to configure the PowerBook with an identical login name and UID/GID, identical network config, and identical .cshrc, everything works fine when logging in to the PowerBook. Unless I figure out what's unique about the G5's configuration, this may not be easy to replicate reliably on other systems. I think my next step is to try to build tcsh with debugging symbols, attach to it with GDB when I log in, and step through the code to see exactly where things go wrong. I may also try to track down exactly which part of the example rc causes the problem. (Presumably it's the section which sets up completion.) This email sent to site_archiver(a)lists.apple.com

1 0

UDF with resource fork, help needed
by mlists＠ugsoft.de 27 May '05

27 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi, The problem is including resource forks. Has anybody out there any experience with this? Thanks for any help and ideas. Uwe _______________________________________________ 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 trying to adjust mkisofs to create UDF images containing Finder file infos and resource forks, etc. Including Finder infos is already done. Converting decomposite unicode HFS+ filenames (special characters) to corresponding UDF, Joliet, Rockridge and HFS filename representation is also done. >From the ECMA 167 and UDF260 specs documentation I cannot understand how the required Extended File Entries and corresponding Stream Directory shall be implemented. E.g., does every UDF directory contain a Stream (sub)Directory with the Extended File Entries, or is the Stream Directory seperated from the UDF dir tree, or what, etc.? All I tried with resource forks resulted in system crashes (You need to restart...) when mounting the UDF image on 10.3.9 ;-( which really makes developing hard. I've also tried to find sample code how to read an UDF CD/image to find out where all the resource fork related pieces live and how they are linked together. E.g. I tried to find the Darwin source for mount_udf with no success. Any idea where I can find this or something similar showing how the resource fork stuff is accessed/created? This email sent to site_archiver(a)lists.apple.com

1 0

Re: fsevents oddities (was Re: EVFILT_VNODES?)
by Hamish Allan 26 May '05

26 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On 26 May 2005, at 23:31, Mark Day wrote: I'll do so. Presumably you'll do so as well? ;) Actually, no! Instead the mdfind in line 14 finds nothing whatsoever. Best wishes, Hamish _______________________________________________ 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… It would be worth filing a bug on that. Because the vnode continued to exist for the entire duration, leaving the vnode's associated name and parent unchanged. Had the vnode been recycled and looked up again, that lookup would have created a new vnode with new parent and name. For a fun experiment, try the following between lines 13 and 14: $ find -x /System -ls &> /dev/null I'll bet that the output in line 15 would be "/Users/hamish/bar/ bar.txt". First of all, it's not just a list of parent nodes. It would be a list of (parent,name) tuples. And if you allowed hard links to directories, it would have to be a list of complete paths from the root. Personally, I would have just a list of parent nodes, then anything which cares can ask for a readdir on those nodes recursively to find the whole path. Note that you'd have to maintain this list both on disk and in memory. (You'd want to get a list of all paths, regardless of whether they were used recently to look up the file, right?) Well, yes and no. No more so that in the current situation, really. What I really want to know is, what does Spotlight do? Because it would appear that it only stores vnodes rather than paths, yet if I perform a Spotlight query first thing after a reboot, it doesn't seem to take any more time to return results. This email sent to site_archiver(a)lists.apple.com

1 0