Re: startx hangs on WinXP, xinit doesn't
Hello, pv. On Mon, 16 Aug 2004 22:29:04 -0700 (PDT) p v wrote: [snip] The only difference in the execution of the two is the environment so I experimented and finally I commented this portion of startx script - if [ x$XAUTHORITY = x ]; then XAUTHORITY=$HOME/.Xauthority export XAUTHORITY fi Once I commented that out I get X to start correctly and the output is the same as it is with xinit on its own. It's been long time since I worked with X - any ideas why XAUTHORITY set makes startx to hang ? thanks Was that really copied verbatim? If so, then there's the problem. For a _test_ of equality, there should be two equal signs, not one. Not true - the syntax is correct, but the action (set XAUTHORITY to a default value if it isn't already set to a non-null string) is inappropriate. The Cygwin X server gets upset if XAUTHORITY doesn't point to a valid authority file containing correct data.
RE: Pathnames for X server app-defaults etc.
Subject: RE: Pathnames for X server app-defaults etc. From: Ruth Ivimey-Cook [EMAIL PROTECTED] To: [EMAIL PROTECTED] Cc: 'Dr.D.J.Picton' [EMAIL PROTECTED] Mime-Version: 1.0 Date: Thu, 25 Mar 2004 16:39:42 + Content-Transfer-Encoding: 7bit X-Spam-Score: 0.0 (/) On Thu, 2004-03-25 at 15:36, Alexander Gottwald wrote: On Thu, 25 Mar 2004, Ruth Ivimey-Cook wrote: The xserver is called XWin.exe. Maybe X is a custom script that starts the xserver with xdmcp parameters. The first lines of the log should contain the commandline parameters. Unfortunately you stripped them It seems to me that we are chasing red herrings here. /usr/X11R6/X is just a symbolic link to /usr/X11R6/Xwin.exe - they are one and the same! I suspect that Ruth's problem is not quite the same as mine. My X server didn't crash; it just failed to find the application default files and fell back on a default for the keyboard map, so that when I typed '#' on my UK keyboard I got '\'. Anyway, the symbolic links fixed the problems. If the same symbolic links don't fix Ruth's problem, she probably needs to look elsewhere. Maybe the first thing to do is to reinstall the X11 packages.
Pathnames for X server app-defaults etc.
Earlier this week, I re-installed Cygwin from scratch. The current version of the X server is 4.3.0-60. I noticed various oddities - in particular, application default files were ignored and the keyboard map was incorrect (for example, '#' produced a '\'). Then I realized that the server now expects to find all the config files under /usr/X11R6/lib/X11. Installing a few symlinks fixed all the problems: cd /usr/X11R6/lib/X11; ln -s /etc/X11 . Now everything works properly. However, it seems to me that something is wrong with the package installation scripts which should either install all the files in /usr/X11R6/lib/X11, or create the relevant symlinks in this directory.
Re: Pathnames for X server app-defaults etc.
From: Dr.D.J.Picton dave at aps5 dot ph dot bham dot ac dot uk To: cygwin-xfree at cygwin dot com Date: Thu, 25 Mar 2004 12:22:01 + (GMT) Subject: Pathnames for X server app-defaults etc. Reply-to: cygwin-xfree at cygwin dot com Reply-to: Dr.D.J.Picton dave at aps5 dot ph dot bham dot ac dot uk I noticed various oddities - in particular, application default files were ignored and the keyboard map was incorrect (for example, '#' produced a '\'). Then I realized that the server now expects to find all the config files under /usr/X11R6/lib/X11. Installing a few symlinks fixed all the problems: cd /usr/X11R6/lib/X11; ln -s /etc/X11 . Sorry - spot the deliberate mistake! That should have been cd /usr/X11R6/lib/X11; ln -s /etc/X11/* .
XFree86 for cygwin installation hanging - a clarification
From: Paul Thomas paulthomas2 at wanadoo dot fr To: cygwin-xfree at cygwin dot com Date: Tue, 28 Oct 2003 06:44:21 +0100 Subject: XFree86 for cygwin installation hanging - a clarification Reply-to: cygwin-xfree at cygwin dot com Hello once more, The problem with the XFree85 install is associated with the XFree-bin-icons package. If I eliminate that from the list of packages, I get a sufficiently complete installation that I can at least run Nedit. XWin still produces a background and nothing else. Yours Paul Thomas There is a fix for the XFree-bin-icons hang (see recent threads on the Cygwin mailing list). This turned out to be an obscure problem in the bash shell, now fixed. If you update the bash shell the bug should go away. It worked for me!
Odd error message from cp and mv when omitting .exe suffix
I have noticed an oddity in the behaviour of the cp command. It gives an incorrect error message when I try to copy an .exe file but refer to it without the .exe suffix: cp calc /usr/local/bin/calc.new This produces the incorrect message: cp: calc and /usr/local/bin/calc.new are the same file. The mv command also shows the bug. -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: Cygwin setup stopped @99%
From: G. Oto genkiotoko at yahoo dot co dot jp To: cygwin at cygwin dot com, cygwin-developers at cygwin dot com Date: Sun, 12 Oct 2003 16:04:38 +0900 (JST) Subject: Cygwin setup stopped @99% I followed the instructions on www.cygwin.com on how to install the current Cygwin, downloaded the setup.exe and ran it. The installation went well until it remained only 1% to the end of the installation. The installation suddenly stopped. I tried 3 times but to no avail. (Please refer to the attached jpeg file for where the installtion stopped. Clicking the cancel button displayed Installation Complete message and nothing more). Can you tell me what to do, please? My pc runs on Windows XP with 1024MB RAM and 120GB HDD. There has been a great deal of correspondence on this issue in the Cygwin mailing list! See posts on cygpath and setup hangs. To complete the installation, start up cygwin and manually run any outstanding .sh scripts in /etc/setup. Then rename them to .sh.done, e.g. mv Xfree86-bin-icons.sh XFree86-bin.icons.sh.done The hang won't occur if you start setup from a .bat file or any type of command window. I circumvent the problem by starting setup.exe from a .bat file: --start of setup.bat-- @echo off C: chdir C:\cygwin setup --end of setup.bat-- (For C:\cygwin, substitute the location of the setup.exe file) -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: Cygwin setup stopped @99%
Date: Tue, 14 Oct 2003 15:51:24 +0100 (BST) From: Dr.D.J.Picton [EMAIL PROTECTED] Subject: Re: Cygwin setup stopped @99% To: [EMAIL PROTECTED] Mime-Version: 1.0 Content-MD5: DsCNL7oYTQKmOyN7mUCpjA== To complete the installation, start up cygwin and manually run any outstanding .sh scripts in /etc/setup. Then rename them to .sh.done, e.g. ^^^ mv Xfree86-bin-icons.sh XFree86-bin.icons.sh.done Sorry - my mistake. You should run outstanding .sh scripts in /etc/postinstall then add the .done suffix. (Alternatively you could re-run setup from a command window or .bat file - this would achieve the same result). -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: SSH connection close when _any_ user logs out of W2K
From: Larry Hall cygwin-lh at cygwin dot com To: Matthew Hilty mhilty at artic dot edu, cygwin at cygwin dot com Date: Tue, 30 Sep 2003 18:53:46 -0400 Subject: Re: SSH connection close when _any_ user logs out of W2K Reply-to: Cygwin List cygwin at cygwin dot com At 05:23 PM 9/30/2003, Matthew Hilty you wrote: Hello, I've noticed on a recent installation of OpenSSH under cygwin (followed http://tech.erdelynet.com/cygwin-sshd.html) that users connected to the server via SSH have their connections closed if any Windows user logs out of the desktop. The SSHD daemon still functions, it just closes active sessions. After this, I can SSH back to the server and my session stays active until I intentionally log out, or another Windows users logs in, then out. Any insights or references would be wonderful; I've been combing mailing lists and usenet and can't find a similar description. If you followed installation instructions for OpenSSH from another site, then you should direct your questions about problems with OpenSSH to that site. tech.erdelynet.com is not cygwin.com and this list, as a result, can't support information for it. My best recommendation, if you'd like someone on this list to entertain the notion of investigating your problem, is to uninstall and reinstall OpenSSH via setup and then to configure it as /usr/share/doc/Cygwin/openssh-*.README suggests. If you still see the same problem, you'll want to visit http://cygwin.com/problems.html first and then follow-up with this list providing the information requested there. FWIW, I just tried a quick test here and I don't see the situation you described. I can't say that this is significant. Only that I can't reproduce the problem with the information given and my Cygwin-supported install. ;-) I just want to flag up the fact that I'm now seeing the same problem, also on a Win2K box. Any processes spawned by sshd are killed when I do a Windows logout. This used not to be the case, so I'll find out whether I see the problem if I revert to the previous version of sshd. I also used the installation procedure in the tech.erdelynet.com website. It boils down to the following - I can't see how it conflicts with anything in the README file: 1. Set the Win2K CYGWIN system environment variable to ntsec tty 2. Install sshd using the setup program 3. Run ssh-host-config -y, also giving 'ntsec tty' in response to the prompt for the CYGWIN value 4. Change some file permissions and ownerships (the significant part being to do: chown system:system /var/log/sshd.log /var/empty /etc/ssh_h*) 5. Run cygrunsrv -S sshd to start the sshd daemon. -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: SSH connection close when _any_ user logs out of W2K
Date: Mon, 13 Oct 2003 15:08:25 +0100 (BST) From: Dr.D.J.Picton [EMAIL PROTECTED] Subject: Re: SSH connection close when _any_ user logs out of W2K From: Larry Hall cygwin-lh at cygwin dot com To: Matthew Hilty mhilty at artic dot edu, cygwin at cygwin dot com Date: Tue, 30 Sep 2003 18:53:46 -0400 Subject: Re: SSH connection close when _any_ user logs out of W2K Reply-to: Cygwin List cygwin at cygwin dot com At 05:23 PM 9/30/2003, Matthew Hilty you wrote: Hello, I've noticed on a recent installation of OpenSSH under cygwin (followed http://tech.erdelynet.com/cygwin-sshd.html) that users connected to the server via SSH have their connections closed if any Windows user logs out of the desktop. The SSHD daemon still functions, it just closes active sessions. After this, I can SSH back to the server and my session stays active until I intentionally log out, or another Windows users logs in, then out. Any insights or references would be wonderful; I've been combing mailing lists and usenet and can't find a similar description. I just want to flag up the fact that I'm now seeing the same problem, also on a Win2K box. Any processes spawned by sshd are killed when I do a Windows logout. This used not to be the case, so I'll find out whether I see the problem if I revert to the previous version of sshd. OK - I've run a couple of quick tests. 1. The bug didn't go away when I reverted to the previous version of sshd. 2. If I run an explicit command using ssh, it doesn't get killed when a Windows user logs out. I suspect that this has something to do with the fact that such processes have no virtual terminal associated with them. For example, the following process will stay around after a logout: ssh hostname bash -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: setup hangs during postinstall
On Thu, Oct 09, 2003 at 06:17:08PM -0500, Brian Ford wrote: #!/bin/bash FOO=`cypath -S` but using #!/bin/sh doesn't. likewise, strace hangs here: #!/bin/bash FOO=`strace -o /tmp/cygpath.strace cypath -S` but not using #!/bin/sh. Neither hang if setup is launched from bash instead of explorer. Hmm. Maybe I missed this point before. I never run setup from explorer. In any event, can you set the CYGWIN_DEBUG=cygpath and maybe do some search and destroy debugging to see precisely where it is hanging? cgf I've done some experimentation with setup with similar results, and can confirm that the hang depends on the environment from which setup is started. I see the cygpath hang (when reinstalling XFree86-bin-icons) in these cases: 1. Clicking on the setup.exe icon in Explorer 2. Clicking on a setup.exe shortcut icon in the desktop I don't see the hang if I start setup from any kind of command window. This includes: 1. A standard Microsoft command window. 2. A command window running bash. 3. An xterm window running bash. I don't see the hang if I start setup from a Cygwin session, or from a standard command window. 4. A .bat file, e.g. @echo off C: chdir C:\cygwin setup -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: Map Backspace to ASCII DEL?
From: Elfyn McBratney elfyn at cygwin dot com To: cygwin at cygwin dot com Date: Thu, 21 Aug 2003 06:01:54 +0100 Subject: Re: Map Backspace to ASCII DEL? References: [EMAIL PROTECTED] Reply-to: cygwin at cygwin dot com Ian Brandt [EMAIL PROTECTED] wrote: I've searched through the mailing list and have seen many posts related to backspace and delete behavior, so my apologies in advance for yet another one, but I can't seem to find the answer I'm looking for in the archives. Currently it seems that the cygwin terminal sends ^H (ASCII BS, 0x08) for backspace, and the VT220 Remove escape sequence (\E[3~, 0x1B5B337E) for Delete. I'd like it to send ^? (ASCII DEL, 0x7F) so that ^H can be used by applications (e.g. emacs). This is how I've always configured other terminal emulators that I've used, and it has worked well. I believe Cygwin just repeats what it gets from Windows. Typically for the console this would changed via keymaps, but I don't see that Cygwin uses this. I don't want to change my mapping in Windows as obviously that would mess up my native environment. Is there a low level way to change the keymap for Cygwin? If not is there a source hack I could implement (and if so where in the source should I look)? `stty erase ^?', IIRC. -- Elfyn No! Stty settings don't change keyboard mappings. All the 'stty erase' setting does is to select the 'character delete' code for use in 'simple' terminal input (e.g. from applications like ftp as opposed to shells which handle line editing themselves). Setting stty erase ^? achieves nothing useful with the default key mappings, because you can't actually generate a delete character! I think that rxvt has the functionality which you require. For example: rxvt -backspacekey ^? -deletekey ^h will give you a session in which the backspace key generates ^h and the delete key generates ^?. (It might then be useful to issue stty -erase ^? so that you can use the backspace key to delete characters in simple terminal input.) In the case of emacs, have you tried running it under X11? In this mode it can distinguish between the backspace key (which it interprets as a 'delete last character' function) and ^h (which calls the help command). -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: Map Backspace to ASCII DEL?
From: Ian Brandt ian at ianbrandt dot com To: Cygwin cygwin at cygwin dot com Date: Thu, 21 Aug 2003 09:16:44 -0400 Subject: Re: Map Backspace to ASCII DEL? References: [EMAIL PROTECTED] Dr.D.J.Picton wrote: [snip] I think that rxvt has the functionality which you require. For example: rxvt -backspacekey ^? -deletekey ^h [snip] rxvt is purely an X application though, no? I'm trying to get this going under the Cygwin console. For X I could use xmodmap which would change the binding before it got to rxvt, xterm, emacs, etc. Actually, rxvt has a native Windows mode (which will run if DISPLAY is unset), although it probably requires some of the X11R6 .dll files to run. So you could try using rxvt in place of a standard console by changing cygwin.bat to do: rxvt -backspacekey ^? -deletekey ^h -e bash --login The only problem I noticed was the small font (which you can change by pressing the '+' key on the keypad in conjunction with the shift key) and the fact that the DISPLAY variable is set (which could be fixed with an /etc/profile hack.) hack). -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: Various shell problems
From: Christopher Faylor cgf-idd at cygwin dot com To: cygwin at cygwin dot com Date: Mon, 18 Aug 2003 14:47:30 -0400 Subject: Re: Various shell problems References: [EMAIL PROTECTED] On Mon, Aug 18, 2003 at 07:37:35PM +0100, Dr.D.J.Picton wrote: [snip] I'll refer to this type of input as 'simple' input. For example, if I create a file using the cat command: cat test the line editing facilities are very limited. [snip] i.e., Cygwin is a Linux-like environment for Windows Just to clarify matters... I wasn't complaining about the limited functionality, only about a change in the 'stty erase' setting which reduced the functionality still further! On a Unix system the choice is really a matter of personal preference, but in the Cygwin environment there are good reasons to use backspace in preference to delete. -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
RE: Is the list hosed?
From: Bill McCormick wpmccormick at covad dot net Date: Tue, 19 Aug 2003 17:59:00 -0500 I'm thinking the list is messed up at this point. I'm getting some very strange responses. Yes - I've had the same problem. The last e-mail i sent re 'various shell problems' got replies back from various sites: [EMAIL PROTECTED] [EMAIL PROTECTED] [EMAIL PROTECTED] The problem may have been caused by a virus, but it can't have been on my system - the mail was sent from a Sun Sparcstation, not a PC. Furthermore, it was created by cutting-and-pasting the previous message from a Netscape window to the compose window - there is no way in which any hidden headers in the original message could have generated the additional recipients. Here are the headers on the message which I sent out. Date: Tue, 19 Aug 2003 10:11:31 +0100 (BST) From: Dr.D.J.Picton [EMAIL PROTECTED] Subject: Re: Various shell problems To: [EMAIL PROTECTED] Cc: [EMAIL PROTECTED], [EMAIL PROTECTED] X-Mailer: dtmail 1.2.1 CDE Version 1.2.1 SunOS 5.6 sun4u sparc Mime-Version: 1.0 Content-Type: TEXT/plain; charset=us-ascii Content-MD5: aclP73oJfVdAqXHOaJ/rug== Content-Length: 900 Here are the headers on one of the messages I got back: Received: from bham.ac.uk (bham.ac.uk [147.188.128.127]) by aps5.ph.bham.ac.uk (8.8.8+Sun/8.8.8) with ESMTP id OAA02047 for [EMAIL PROTECTED]; Wed, 20 Aug 2003 14:46:38 +0100 (BST) From: [EMAIL PROTECTED] Received: from mail.virginia.edu ([128.143.2.9]) by bham.ac.uk with smtp (Exim 3.36 #1) id 19pTHI-00053Y-00 for [EMAIL PROTECTED]; Wed, 20 Aug 2003 14:45:32 +0100 To: [EMAIL PROTECTED] Subject: Majordomo results: Re: Various shell problems Reply-To: [EMAIL PROTECTED] Date: Wed, 20 Aug 2003 9:45:30 EDT Sender: [EMAIL PROTECTED] Message-Id: [EMAIL PROTECTED] x-scan-bham: no Content-Length: 2890 -- From: Christopher Faylor cgf-idd at cygwin dot com Command '' not recognized. To: cygwin at cygwin dot com -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: Various shell problems
From: John Morrison john dot r dot morrison at ntlworld dot com To: cygwin at cygwin dot com Date: Sun, 17 Aug 2003 10:18:56 +0100 Subject: RE: Various shell problems From: Dr.D.J.Picton Hi :) Thanks for the feedback, [snip] Unfortunately, I did notice problems with /etc/profile: Firstly, it sets stty erase ^? In my view the correct setting is definitely ^h (backspace). *GRIN* before we get into this discussion, I'd like to note the in my view :) To quote: Both keys essentially do the same thing, with one major difference. The backspace key deletes to the left of the cursor, and the delete key deletes to the right of the cursor. See below. I wasn't talking about bash shell input which does indeed work as described! Setting the erase character to ^? (delete) causes problems when 'normal' terminal i/o is used. What's a 'normal' terminal? Do you mean the one you start from the cygwin shortcut? If so, it works for me; what version of bash are you running? (it only gets set for bash atm) I'm using: GNU bash, version 2.05b.0(9)-release (i686-pc-cygwin) Unfortunately there has been a slight misunderstanding here. I did mean the cygwin shortcut, and I am using the above version of bash. However... By 'normal' input I meant the 'dumb' input to programs which issue simple terminal reads, e.g. cat or ftp (as opposed to shells which do complex line editing.) I'll refer to this type of input as 'simple' input. For example, if I create a file using the cat command: cat test the line editing facilities are very limited. I find I can't reposition the cursor using the arrow keys. This moves the cursor, but the cursor key escape sequences are stored in the file! The only way to correct the input is to backspace from the end of the line, and the backspace key is the logical choice - I'm removing the character to the left of the cursor. This works if the stty erase character is ^h. However, if it's been set to ^?, things tend to break: 1. In a window started by the cygwin shortcut, I can't backspace input in simple terminal reads. The delete key does nothing, and the backspace key actually generates a ^h character! 2. In an xterm window without the resource *VT100*deleteIsDel: true Bash input works as advertised, but setting stty erase ^? completely breaks character deletion in 'simple' input. The backspace key generates an actual ^h character, and the delete key produces an escape sequence ... 3. In an xterm window with the resource *VT100*deleteIsDel: true Again, there's no problem with bash input. I can now set stty erase ^? and use the delete key as the character erase key in simple input. But this is all very counter-intuitive, for the reasons given above. Another thing I've noticed is that stty echoctl is set in an xterm session, but it doesn't actually seem to do anything. I would expect control characters to be echoed in 'caret' notation in simple input (e.g. ^a for control-A), and on other systems I find this very useful because it lets me know when I'm entering control characters. Is this a known limitation of the terminal emulation under Cygwin? inetutils (of which ftp is part) is at version 1.4.2 on GNU, cygwin's ftp is (GNU inetutils) 1.3.2. Could you try this version and, if it fixes the bug, lobby to get cygwin updated? I'll look into this when I have the time (which isn't right now!) In the meantime, the /etc/profile kludge will keep the users happy. But of cause, it's *your* /etc/profile now, feel free to hack it :) Thanks again for taking the time, J. -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: startx and ssh
Mail-Followup-To: [EMAIL PROTECTED] Delivered-To: mailing list [EMAIL PROTECTED] To: [EMAIL PROTECTED] Subject: Re: startx and ssh Date: Sat, 16 Aug 2003 03:39:02 -0400 From: John P. Rouillard [EMAIL PROTECTED] x-scan-bham: no Even better, don't leave the X server wide open for everybody to screw with. I'm puzzled by this assertion. The default is to allow access only from the local machine, and in my view this is secure enough for most people! The XFree-86 server includes the security extension (see xdpyinfo) so you should be able to generate a proper X authentication token for the display. Something like: xauth generate :0 . or xauth add $DISPLAY . `mcookie` added to the top of your ~/.xinitrc should do the trick. I would like to add a note of caution here. For some reason, Magic Cookie authorization only works for truly local connections i.e. :0, but the server throws up a 'Protocol Not Specified' error for network displays, e.g. 127.0.0.1:0. If you set up the server with Magic Cookie authorization as the only means of access, you will have to make sure that DISPLAY is set to :0, and change any scripts which specify -display 127.0.0.1:0. -- rouilj John Rouillard === My employers don't acknowledge my existence much less my opinions.
Re: Various shell problems
Date: Fri, 15 Aug 2003 14:23:06 -0400 From: Larry Hall [EMAIL PROTECTED] To: Dr.D.J.Picton [EMAIL PROTECTED] CC: [EMAIL PROTECTED] Subject: Re: Various shell problems Content-Transfer-Encoding: 7bit x-scan-bham: no Dr.D.J.Picton wrote: 3. /bin/sh sources /etc/profile when called from ftp.exe If SHELL isn't exported or is set to /bin/sh, ftp.exe uses the /bin/sh shell for shell escapes. Unfortunately the working directory of the shell is then always set to the user's home directory. Now I know why - for some reason, /bin/sh thinks it's a login shell and sources the profile. Here is my fix to /etc/profile to circumvent the problem. All the code is skipped if it detects that the profile has already been run, and argument zero is set to -sh: At the start of the file: # fix for ftp problem ... if [ -z $PROFILE_ALREADY_DONE -o $0 != -sh ]; then export PROFILE_ALREADY_DONE=1 At the end of the file: fi -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Various shell problems
1. I downloaded the new version of base-files (2.0-2) today. I note that the new config files now go into the /etc/defaults directory, which is a good idea! Although it does have one disadvantage - if there are problems in the new files, many existing users won't see them. Unfortunately, I did notice problems with /etc/profile: Firstly, it sets stty erase ^? In my view the correct setting is definitely ^h (backspace). Setting the erase character to ^? (delete) causes problems when 'normal' terminal i/o is used. The delete key doesn't seem to work in a Windows command window, and a resource (deleteIsDel) has to be set to make it work in an xterm window. Secondly, the section which checks whether the user's passwd entry has been set up uses nonstandard operators (==) which aren't supported by all sh-like shells. This is what happens when profile is sourced by /bin/sh: [: ==: unknown operand I changed == to =, and now all is well. 2. SHELL environment variable problem. For some reason, the bash shell tends to drop the SHELL variable from the environment, causing a variety of problems. I fixed this with a new startup script (/etc/bash.bash_env) which issues 'export SHELL', and by contriving (via the BASH_ENV environment variable and .bashrc scripts) that all bash shells will use the script. 3. /bin/sh sources /etc/profile when called from ftp.exe If SHELL isn't exported or is set to /bin/sh, ftp.exe uses the /bin/sh shell for shell escapes. Unfortunately the working directory of the shell is then always set to the user's home directory. Now I know why - for some reason, /bin/sh thinks it's a login shell and sources the profile. This bug seems to be specific to ftp - I haven't seen it with sftp. -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: Various shell problems
Date: Fri, 15 Aug 2003 14:23:06 -0400 From: Larry Hall [EMAIL PROTECTED] To: Dr.D.J.Picton [EMAIL PROTECTED] CC: [EMAIL PROTECTED] Subject: Re: Various shell problems 2. SHELL environment variable problem. For some reason, the bash shell tends to drop the SHELL variable from the environment, causing a variety of problems. I fixed this with a new startup script (/etc/bash.bash_env) which issues 'export SHELL', and by contriving (via the BASH_ENV environment variable and .bashrc scripts) that all bash shells will use the script. This is a change in bash behavior. It's intentional. It's not something new though. It's at least a year old. You can check the email archives for the bash announcement on this issue and/or the subsequent reports of bugs from this behavior change. Thank you for clarifying this. I obviously need to do my homework more thoroughly in the future! The ftp problem occurs because the /bin/sh shell is called with $0='-sh' (whereas other shells have $0='+bash' etc and therefore presumably don't regard themselves as login shells). I'll modify profile to detect the situation and avoid doing cd $HOME. -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
XFree86-start-menu-icons path problem
I went into the start menu, clicked Programs/Cygwin-XFree86/editors/emacs. No response. (I had started the X11 server!) So I looked at the properties of the emacs shortcut, and simulated the operation in a command window: cd c:\CYGWIN\bin c\:\CYGWIN\usr\X11R6\bin\run.exe emacs This failed to work. Typing just 'emacs' revealed the problem - the system could find the /bin DLL's but not the ones in /usr/X11R6/bin. Appending this directory to the path then made the start menu icons work: set path=%path%;c\:CYGWIN\usr\X11R6\bin A similar change to the system path enabled the start menu icons to work. But is it a recommended practice to include Cygwin directories in the standard system path? (It makes it easier to launch Cygwin applications from a Windows environment, but I'm worried about potential side effects). In my view it would be preferable for the path setting to be handled by the shortcut.
Re: startx and ssh
Date: Mon, 11 Aug 2003 12:55:05 -0600 I have been trying to de-bug a problem with startx. When I tried to run the file, I would get a server window, no clock and no terminal windows. The command-line errors were: .. Xlib: connection to 0.0 refused by server Xlib: No Protocol specified These error messages would repeat until I hit ^Z from the console command line. I found that the script would fail if a .Xauthority file existed in my home directory. If I removed the file, startx ran without errors; the server, the clock and the terminal windows all opened successfully when there was no .Xauthority file in my home directory. I found that the .Xauthority file was being created when I logged onto my Cygwin machine from another machine using ssh with the same login name. I am proposing a fix for this problem: In /usr/X11R6/bin/startx, add the following code before the block that exports the XAUTHORITY variable: # remove $HOME/.Xauthority if it exists if [ -f $HOME/.Xauthority ]; then rm $HOME/.Xauthority fi This code just removes .Xauthority if it exists. If .Xauthority doesn't exist, it does nothing. Does anyone have any feedback? I've seen the problem, and in my view there's a better solution. Remove the code which sets XAUTHORITY. (I think the problem occurs because XWin assumes -auth $XAUTHORITY if the variable is set, then gets upset because it can't find the authorization key for your display.)
Re: mouse right-click problems on Sun
From: Bernard Disselborg [EMAIL PROTECTED] To: [EMAIL PROTECTED] I'm also using a 3 button mouse, and experience this problem both on (Sparc) Solaris 2.6 an 8. I've also installed the latest Logitec Mouseware drivers, but this did not help. One thing to try - start up xev, move the pointer into the window, and see how it responds to mouse buttons. The mouse I'm using is really a 2 button mouse with a central track wheel which the X server regards as the middle button. Turning the wheel gives a further two 'buttons'. Left button - button 1 (i.e. left) Depress trackwheel- button 2 (i.e. middle) Right key - button 3 (i.e. right) Push trackwheel away - button 4 (press+release) Pull trackwheel towards - button 5 (press+release) What response does xev give to a right mouse press? If it's an incorrect button (e.g. button 5), you'll be able to fix things up using xmodmap.
Xwin remote calls to Solaris machines/ font server bug
I have two observations to make concerning the use of 'XWin -query' to Solaris machines. 1. Firstly, the FAQ states that Solaris can't cope with 24-bit colour depths. This may have been true for older releases, but all recent versions of Solaris (and recent Sun hardware) support this depth! 2. The FAQ refers to the 'hang' when you attempt to contact a remote Solaris host using XWin -query e.g. Xwin -query sunhost The cure is to specify a Solaris font server on the command line, e.g. XWin -query sunhost -fp tcp/sunhost:7100 However, this shouldn't be necessary for recent Solaris releases. The XSetup script on the Sun machine automatically adds the font server to the font path, before the dtlogin window appears. The real reason for the problem is a bug in XWin, which 'hangs' if I try to mix local fonts with a Solaris font server. The following procedures work OK: 1. Start up an XWin server with normal local fonts, then replace them with a Solaris font server: xset fp= tcp/sunhost:7100 2. Start up an Xwin server with -fp tcp/sunhost:7100 on the command line. 3. Add a second Solaris font server when the fontpath already contains one. The following procedures cause a 'hang': 1. Start up an XWin server with normal local fonts, then append a Solaris font server: xset fp+ tcp/sunhost:7100 2. Start up an XWin server with -fp tcp/sunhost:7100 on the command line, then try to append a local font to the path: xset fp+ /usr/X11R6/lib/X11/fonts/misc
Re: HELP, xinit failure
From: Rodrigo Medina [EMAIL PROTECTED] To: [EMAIL PROTECTED] Subject: HELP, xinit failure Date: Sat, 9 Aug 2003 08:48:02 -0400 Hello, After updating XFree86, xinit no longer works. A WINDOWS alert window appears telling that LIBICE.DLL cannot be found. Can somebody tell me if that library is a windows library or if it is a cygwin/xfree library? In the last case, in what package is it included? It's in XFree86-lib-compat - see below. Thanks in advance. Rodrigo Medina. I suspect you have two problems: 1. Your X11 binaries haven't all been updated. I suggest that you check the date of the xinit binary: ls -l /usr/X11R6/bin/xinit.exe If it isn't 1 August, you don't have the current version. To fix this, close down all cygwin processes then launch setup. Reinstall the latest version of the XFree86-bin package. 2. You may also lack the compatibility package, which ensures that 'old' (i.e. pre-upgrade) X11 binaries will continue to work. From setup, select the following package for installation: XFree86-lib-compat.
4.3.0-1 XFree86-etc package problem
Apologies if this has been reported before, but there seems to be a basic problem with this version. I've tried two download sites and in both cases the files are installed in the wrong directory - /etc/X11/X11 instead of /etc/X11. This caused some unpleasant symptoms which mostly went away if the files were moved back into /etc/X11. Unfortunately xterm still didn't work properly because the XTerm and XTerm-color files have gone missing from the app-defaults directory. I've reverted to 4.2.0-1 for the time being.