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

Tiger 1024-char limits on tcsh shell command length ... maybe?
by OpenMacNews 03 May '05

03 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com hi all, ################ ## details: # (i think i'm getting closer to the right list to ask abt this on ... peeking in darwin8's tcsh src/dpkg/control i note: Package: tcsh Maintainer: Darwin Developers <darwin-development(a)public.lists.apple.com> it looks like this may have be the right(er) place for this Q) as per my original post to darwin-userlevel: <http://lists.apple.com/archives/darwin-userlevel/2005/May/msg00008.html> (1a) copy-n-pasting to a bash prompt --> executes OK (1b) copy-n-pasting to an Apple native tcsh prompt --> executes OK (1c) copy-n-pasting to a custom-built (w/ -DBUFSIZE=8192 -DMAXPATHLEN=4096) tcsh prompt --> executes OK for a ./configure string > 1024 characters in length, as per example in my prior post: (2a) copy-n-pasting to a bash prompt --> executes OK (2b) copy-n-pasting to an Apple native tcsh prompt --> system bells, 1 for each char > 1024 (2c) copy-n-pasting to a custom-built (w/ -DBUFSIZE=8192 -DMAXPATHLEN=4096) tcsh prompt --> system bells, 1 for each char > 1024 HOWEVER, if to a (any) tcsh prompt i copy-n-paste the FIRST 1024 characters, of the cmd string then copy-n-paste the NEXT 1024 characters, etc. and *THEN* hit return to execute -- executes OK !!! huh ?? how come no cmd line length limit in this case? justin walker @ darwin-userlevel SWAG'd: could be, but i'm in over my head as to kernel code behavior and 'raw mode' ... also, given darwin8 pkg info for its tsch-46, for the moment, i'm stumped and confused. thoughts/suggestions? cheers, richard _______________________________________________ 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 hit a problem with Tiger's tcsh shell behavior -- specifically, hitting a limit on command line length, but with a bit of a curious twist ... ################ ## short descr: # copy-n-paste of long (>1024 char) strings to tcsh prompt worked in 10.3.9, but is broken in 10.4. i hit this problem when trying to copy-n-paste a very long "configure" string during a php504 (re)build on OSX 10.4. for a ./configure string <= 1024 characters in length, as per example in my prior post <http://lists.apple.com/archives/darwin-userlevel/2005/May/msg00008.html> (sorry for the thread break ...): now, iiuc, there's a 'known' limit of 1024 characters for many unix shell commands. "A Scientific Wild-ass guess: there may be a difference between how the shells handle going in and out of 'readline' mode, which could account for the difference you are seeing. The fact that you get bells indicates that the tty code in the kernel has hit a limit, which may not be affected by increasing the limit for which your shell has been compiled. Since it works when you paste and paste again, I am guessing that tcsh is in "raw" mode, and the timing difference between one paste and two pastes is sufficient to avoid the bells in the latter case." Vendor: Christos Zoulas Version: 6.10.00 URL: ftp://ftp.astron.com/pub/tcsh Description: TENEX C-Shell Build-Depends: build-base, gnudiff i've already chatted whith Christos Zoulas directly & he simply verified that he does NOT see the problem on Linux, and that -DBUFSIZE=8192 -DMAXPATHLEN=4096 should've done the trick ... but no dice. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Getting a mach_header from dlopen
by Peter O'Gorman 03 May '05

03 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 Spencer Nielsen wrote: | Does anybody know offhand a good way to get a code image's mach_header | (for use with getsectdatafromheader) of a specific file? In 10.3 I | used the structure that dlopen returned (dlstatus) and it had a pointer | right in there. However, things have changed in 10.4 and dlopen now | returns a different structure (dlopen_handle). The structure returned | now contains the NSModule for the file among other things. I tried | running through all the loaded images with _dyld_get_image_header and | comparing the names (using _dyld_get_image_name) with the | NSNameOfModule() called on the NSModule in the structure returned by | dlopen but NSNameOfModule() always returns nil. I know that depending | on what is returned by dlopen is probably not the best of strategies | (perhaps that is why the man page still documents it as returning | void*). All I need to be able to do it get the mach_header of code | images that I am loading in with dlopen. Thanks! | Well, you really shouldn't be making any assumptions about what dlopen() returns, it already broke for you 10.3->10.4 and it may break even on minor OS updates because it is just an implementation detail. Since NSAddImage returns a mach_header, I'd suggest using NSAddImage with the path you used to dlopen and NSADDIMAGE_OPTION_RETURN_ONLY_IF_LOADED | NSADDIMAGE_OPTION_RETURN_ON_ERROR, but I am not quite sure what you are trying to do. Please file an enhancement request with Apple for an api to get information on the handle returned by dlopen. I know that they are trying to reduce the number of dyld apis and standardize on dlopen et.al. where possible, but there are a bunch of things that require a mach_header, so some method of getting a mach_header from a dlopen handle is required. Peter - -- Peter O'Gorman - http://www.pogma.com -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) iQCVAwUBQnbJe7iDAg3OZTLPAQKREwQAr7v95LWZQEW6+vEuoOnb3tpyzIEPFvb8 jgRLFkPsNeVl5lR0gf1L2UG+S6HiV5mXIKVA2yvlzhln7OsG1uykXCWQfFnF6rN0 IpeKIJnK65Yq1Gd+nA7PPhdNZnksPsijkvfdaEl8rxy9A/oUbIlSjDMRkuCYEjbR YOq154YPTL4= =f+36 -----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

Getting a mach_header from dlopen
by Spencer Nielsen 02 May '05

02 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:mime-version:content-transfer-encoding:message-id:content-type:to:from:subject:date:x-mailer; b=IlW9M++Ix5ZpU5VydxeytW12h1j1lk3D3HECq0JUgr9GgbD9x5yJTyrMs7dHQQ8RgDOiRQA3UBl+g4nlZMXyNlNggaBD5UerGXl0IC9Wfek75U58NK+tXyC/b0FtFuLKXhTUu3Ag/Dsr9fWi+DnHs001ew+29lOOvABN6Lg3vso= Spencer Nielsen _______________________________________________ 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… Does anybody know offhand a good way to get a code image's mach_header (for use with getsectdatafromheader) of a specific file? In 10.3 I used the structure that dlopen returned (dlstatus) and it had a pointer right in there. However, things have changed in 10.4 and dlopen now returns a different structure (dlopen_handle). The structure returned now contains the NSModule for the file among other things. I tried running through all the loaded images with _dyld_get_image_header and comparing the names (using _dyld_get_image_name) with the NSNameOfModule() called on the NSModule in the structure returned by dlopen but NSNameOfModule() always returns nil. I know that depending on what is returned by dlopen is probably not the best of strategies (perhaps that is why the man page still documents it as returning void*). All I need to be able to do it get the mach_header of code images that I am loading in with dlopen. Thanks! This email sent to site_archiver(a)lists.apple.com

1 0

Re: DiskArbitration API
by Dan Markarian 02 May '05

02 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi Jesper, Dan On 2 May 2005, at 5:52 AM, Jesper Papmehl wrote: Hello! Thanks for any help! /Jesper Papmehl Propellerheads _______________________________________________ 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… Disk Arbitration SPIs do continue to work on Tiger. What do you do with the port you are handed back in DiskArbStart()? Do you add it do your run loop? You must add it to your run loop in order to service it and have your notifications fire. You can also use DiskArbMsgLoop() to service it as an alternative. You may ask how some applications do work on Panther even with the oversight. The port from the SPI is a global port, for historical reasons. You may get lucky in case someone else in your application does do the right thing by adding it to the main run loop. It was probably File Manager. File Manager has since moved onto the new APIs. You must service the port you are handed back to be sure to receive notifications. The same is true of the new APIs, which are also present on Panther, and most other notification APIs, as a general rule. You can take a look at the disktool.c sample code, in the DiskArbitration project, to see how to service that port with a run loop on the old SPI. You are welcome to use the new API as well. You may sign up with DARegisterDiskPeekCallback() and claim your disk in that callback with DADiskClaim(). We have an app that reads (removable) disks that uses a file system that Mac OS X doesn't recognize. We have been using the DiskArbitration API to make the OS ignore the disks as they are inserted, to prevent the "disk cannot be read" dialog from appearing. This has been working fine, until Tiger, where it stopped working because the DiskArbitration API has been completely redesigned. (Not that I'm complaining, I know that we have been using a private, unsupported API.) I have not been able to find any documentation for the new API (apart from the brief comments in the headers), and I'm having a bit of trouble getting started. What I want to do is to get notified whenever a disk is about to be mounted, and then tell the OS to ignore the disk. Does anyone know how to do this using the Tiger version of DiskArbitration? This email sent to site_archiver(a)lists.apple.com

1 0

Re: sysent and KEXT
by Justin Walker 02 May '05

02 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 2, 2005, at 6:47, Marco Bambini wrote: Hi All, So, if I write something like this in my KEXT: #include <sys/syscall.h> then I have a compilation error: "sys/syscall.h: No such file or directory" How can I solve this problem? Or better, is there another way to patch a system call inside a kext? There is no good way to patch system calls. Regards, Justin -- Justin C. Walker, Curmudgeon at Large Institute for General Semantics ----------- My wife 'n kids 'n dog are gone, I can't get Jesus on the phone, But Ol' Milwaukee's Best is my best friend. ----------- _______________________________________________ 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 need to patch a system call using the sysent routine in a KEXT. No, you don't. It's unsupported, and if this is intended as a product, your product will continually break as new releases come out. Think of another way to do what you want. As far as I know, for a kext, I should only use headers from Kernel.framework and sys/syscall.h is a userland header... Yup; kexts are built with headers only from the kernel framework; the build does not look in "the usual places". This email sent to site_archiver(a)lists.apple.com

1 0

Re: sysent and KEXT
by Chris Ridd 02 May '05

02 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Microsoft-Entourage/11.1.0.040913 On 2/5/05 2:47 pm, Marco Bambini <marco(a)vampiresoft.com> wrote: > Hi All, > > I need to patch a system call using the sysent routine in a KEXT. > As far as I know, for a kext, I should only use headers from > Kernel.framework > and sys/syscall.h is a userland header... > > So, if I write something like this in my KEXT: > #include <sys/syscall.h> > > then I have a compilation error: > "sys/syscall.h: No such file or directory" > > How can I solve this problem? > Or better, is there another way to patch a system call inside a kext? There's an example of doing this at: <http://landonf.bikemonkey.org/code/macosx/ptrace_deny_attach.20041010201303 .11809.mojo.html> Hope it helps! Cheers, Chris _______________________________________________ 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

sysent and KEXT
by Marco Bambini 02 May '05

02 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi All, So, if I write something like this in my KEXT: #include <sys/syscall.h> then I have a compilation error: "sys/syscall.h: No such file or directory" How can I solve this problem? Or better, is there another way to patch a system call inside a kext? Thanks a lot for your help. Regards, Marco Bambini _______________________________________________ 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 need to patch a system call using the sysent routine in a KEXT. As far as I know, for a kext, I should only use headers from Kernel.framework and sys/syscall.h is a userland header... This email sent to site_archiver(a)lists.apple.com

1 0

Re: Tiger and libreadline
by chet＠cwru.edu 02 May '05

02 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Read-receipt-to: chet.ramey(a)case.edu > On 01/05/2005, at 6:56 PM, Martin Costabel wrote: > >> Many GNU packages rely on libreadline. > >> It seems that in Darwin there is a BSD libedit, which is symlinked > >> as libreadline and has a corresponding readline.h. This breaks > >> Octave and probably other builds. > >> > > > > Yes, this is annoying, but it fits into the pattern of Apple first > > putting a broken version of a new open source package into OSX > > (there are quite a few historical examples for this). A correct > > version can probably be expected for the OSX release "Sabre-tooth > > cat" :-) > > The Tiger version is neither broken nor 'incorrect'. It does what it > says on the tin. It is however the BSD version not the GNU version. It is, however, incompatible. There is no `BSD version' versus `GNU version'. The BSD libedit resembles readline in that they both provide emacs and vi editing modes, but there is neither source nor object compatibility. There has been some attempt to provide both, but it's not there yet (try key binding or multibyte character support). Chet -- ``The lyf so short, the craft so long to lerne.'' - Chaucer ( ``Discere est Dolere'' -- chet ) Live...Laugh...Love Chet Ramey, ITS, CWRU chet(a)case.edu http://tiswww.tis.cwru.edu/~chet/ _______________________________________________ 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: Tiger and libreadline
by Chris Zubrzycki 02 May '05

02 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 1, 2005, at 11:27 PM, Bill Northcott wrote: Fink has a patch line -chris zubrzycki - -- PGP public key: http://homepage.mac.com/beren/publickey.txt ID: 0xA2ABC070 Fingerprint: 26B0 BA6B A409 FA83 42B3 1688 FBF9 8232 A2AB C070 ======================================================== "Sadly, text alone cannot convey the depths of my sarcasm." _______________________________________________ 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… perl -pi -e 's|LIBREADLINE=\"-lreadline|LIBREADLINE=\"-L%p/lib - lreadline|g' mkoctfile in its octave package description for Tiger. Replace %p by /usr/ local in your case. Actually by default /usr/local/lib should be searched before /usr/lib and /usr/local/include before /usr/ include, so this shouldn't even be necessary in your case. At link time -lreadline will pick up the GNU version, but the problem is the in the configuration file. Also I believe I believe in fixing once and fixing properly, not maintaining a massive and growing database of patches like Fink. Most, if not all patches go upstream from fink. If upstream source doesn't want to apply it, we still need the patch. For a good example of "fixing once and fixing properly" see KDE. This email sent to site_archiver(a)lists.apple.com

1 0

Re: pty-redir: Slave, Master, and how it all works
by Pelle Johansson 02 May '05

02 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; sv-SE; rv:1.6) Gecko/20040113 John Davidorff Pell wrote / skrev: _______________________________________________ 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 want to start a shell on a pty, then attach to the tty (using screen (1), for example), pop, have a shell. Screen(1) has the wonderful ability to attach to a specified tty! :-) Yay! Ok, that side is done, since I use screen(1) already. pty-redir(1) is intended to do the other side. The code I'm using is posted here [1]. The virgin 0.1 release exhibits the same problem, and 0.2 is debian specific. The code at [1] is a merger of the two. :-) You seem a bit confused about how ptys work. Shells don't run on ptys, they run on ttyps. Terminal emulators such as screen and remote terminal servers such as sshd run on ptys. screen also has the ability to attach to another tty as a relay, but that's mostly meant to talk to modems. Is it a reattachable shell you want? (similar to screen). That is done by starting a daemon that attaches to a pty, and starting a shell on the tty. You then use whatever method you wish to connect to the daemon (most likely a socket, telnet protocol might be appropriate) which relays the data to the pty. I guess you could also open a second pty, attach to the corresponding tty with screen, and relay the data between the ptys. It would then be similar to talking through a modem connection. One could imagine using pty-redir for the lat option. You just have to make it start some other program on another pty that starts the final one on the matching tty. -- Pelle Johansson <morth(a)morth.org> This email sent to site_archiver(a)lists.apple.com

1 0