Re: [Cooker] Some problems with terminal-server
On Thu, 12 Sep 2002, Buchan Milne wrote: Yesterday, I spent the day, and a large portion of the evening setting up a server for a school, running 9.0rc2 (with updates from cooker) as: 1)Domain controller and print server 2)SNF 3)Mail gateway 4)Terminal server (I know this isn't the best idea, but cash is limited - the whole project including all new hardware is less than one of the new firewall appliances, and we can only afford one machine until we can prove the value of diskless clients). I had some issues, which currently result in the clients not booting. I will have one opportunity to fix this, probably on Saturday morning (connectivity over a dialup also, so limited email/googling etc). Unfortunately I didn't have time to investigate too much, but here we go: The clients booted their kernel, I think managed the pivot ok, then started up the init scripts, and just died there. One file listed as missing was /etc/mtab, gprintf was not available, and then I got dumped into a root shell saying I had to run fsck on some device, which shouldn't have been mounted, and didn't exist in /dev on the machine. /dev/ appeared to be very empty (about 40 files I think). If I remember correctly, /dev/tty0 was missing (sorry, it was quite late by that stage). If I want to try and fix these issues, what kind of things am I going to need to do? Run msec 2 and all should be well. The terminal-server package exports the server's root file system with all_squash set, which means that root on the workstation is effectively an anonymous user on the server. Therefore, an anonymous user must be able to read all the files necessary in order to boot a workstation. This includes things like /etc/rc.d/init.d/functions. One annoyance is that it seems impossible to have terminal-server and snf configure dhcp, but terminal-server doesn't make allowance for a dynamic range or WINS IPs (both of which I need for the +- 10 windows desktops they currently have), if I were to add another interface to the machine I don't see how I would be able to run two dhcpd's (yet, I could hack a bit), and the SNF config also trashes everything the terminal server setup. terminal-server provides an example DHCPD configuration file, which is made active only if you do not have a DHCPD configuration file at the time of installing terminal-server. The *only* thing that you need to add to a working DHCPD config file in order to turn it into a DHCPD config file that will work with terminal-server clients is the single line: include /etc/dhcpd.conf.etherboot.include; Absolutely everything else is automatic. Please note: I am talking about the terminal-server package as originally built by me; it is possible that the official Mandrake package works slightly differently. My package can be downloaded from www.fensystems.co.uk/RPMS.fensys/. (The need for a configuration api which would make these two tools live together is made evident by this setup). Not really; as I said, you need to add only one line to a working DHCPD config file in order to get it to serve terminal-server clients. [ However, while we're on the subject of configuration, you might want to take a look at the pre-alpha release of fenconfig, which is also available from www.fensystems.co.uk/RPMS.fensys. ] Michael
Re: [Cooker] Some problems with terminal-server
On Thu, 12 Sep 2002, Buchan Milne wrote: Yesterday, I spent the day, and a large portion of the evening setting up a server for a school, running 9.0rc2 (with updates from cooker) as: 1)Domain controller and print server 2)SNF 3)Mail gateway 4)Terminal server (I know this isn't the best idea, but cash is limited - the whole project including all new hardware is less than one of the new firewall appliances, and we can only afford one machine until we can prove the value of diskless clients). I had some issues, which currently result in the clients not booting. I will have one opportunity to fix this, probably on Saturday morning (connectivity over a dialup also, so limited email/googling etc). Unfortunately I didn't have time to investigate too much, but here we go: The clients booted their kernel, I think managed the pivot ok, then started up the init scripts, and just died there. One file listed as missing was /etc/mtab, gprintf was not available, and then I got dumped into a root shell saying I had to run fsck on some device, which shouldn't have been mounted, and didn't exist in /dev on the machine. /dev/ appeared to be very empty (about 40 files I think). If I remember correctly, /dev/tty0 was missing (sorry, it was quite late by that stage). If I want to try and fix these issues, what kind of things am I going to need to do? Run msec 2 and all should be well. The terminal-server package exports the server's root file system with all_squash set, which means that root on the workstation is effectively an anonymous user on the server. Therefore, an anonymous user must be able to read all the files necessary in order to boot a workstation. This includes things like /etc/rc.d/init.d/functions. One annoyance is that it seems impossible to have terminal-server and snf configure dhcp, but terminal-server doesn't make allowance for a dynamic range or WINS IPs (both of which I need for the +- 10 windows desktops they currently have), if I were to add another interface to the machine I don't see how I would be able to run two dhcpd's (yet, I could hack a bit), and the SNF config also trashes everything the terminal server setup. terminal-server provides an example DHCPD configuration file, which is made active only if you do not have a DHCPD configuration file at the time of installing terminal-server. The *only* thing that you need to add to a working DHCPD config file in order to turn it into a DHCPD config file that will work with terminal-server clients is the single line: include /etc/dhcpd.conf.etherboot.include; Absolutely everything else is automatic. Please note: I am talking about the terminal-server package as originally built by me; it is possible that the official Mandrake package works slightly differently. My package can be downloaded from www.fensystems.co.uk/RPMS.fensys/. (The need for a configuration api which would make these two tools live together is made evident by this setup). Not really; as I said, you need to add only one line to a working DHCPD config file in order to get it to serve terminal-server clients. [ However, while we're on the subject of configuration, you might want to take a look at the pre-alpha release of fenconfig, which is also available from www.fensystems.co.uk/RPMS.fensys. ] Michael Brown http://www.fensystems.co.uk -- Fen Systems: Linux made easy for schools
Re: [Cooker] terminal-server
On Tue, 10 Sep 2002, Per Øyvind Karlsen wrote: the terminal-server has ALOT of issues with the init scripts, especially when you have a raidtab on the server *sigh*, I went back to ltsp Thanks for the positive feedback - I really appreciate comments like that. You need to do two things: 1. Run msec 2 on the server, so that files like /etc/rc.d/init.d/functions are available when exported via NFS. 2. Create an empty file /etc/raidtab$$CLIENT$$. This will mask your server's raidtab from the workstations, in the same way that /etc/fstab$$CLIENT$$ masks your server's fstab from the workstations. I'll add (2) to the package. (1) is already documented in the package description. Michael
Re: [Cooker] RC1: Terminal server install routine broken
On Tue, 3 Sep 2002, Stew Benedict wrote: Yes. I used the GUI. This is the contents of /etc/dhcpd.conf: #dhcpd.conf - generated by drakTermServ ddns-update-style none; # Long leases (48 hours) default-lease-time 172800; max-lease-time 172800; # Include Etherboot definitions and defaults include /etc/dhcpd.conf.etherboot.include; # Network-specific section subnet 192.168.0.0 netmask 255.255.255.0 { option routers 202.0.35.1; option subnet-mask 255.255.255.0; option broadcast-address 192.168.0.255; option domain-name ext.dev-zone.org; option domain-name-servers 203.96.152.4, 127.0.0.1; } # Include client machine configurations include /etc/dhcpd.conf.etherboot.clients; To be honest I don't know what the setting routers means. I guess it got that from my static IP address, which ends in another number, not 1. You've got a few major problems there: option routers is presumably picked up from your default gateway setting. I'm guessing you've got two network cards: one internal and one external (or possibly a dial-up link etc in place of the external). If so, then you need to set option routers 192.168.0.x; so that clients on the internal network use your box as the default gateway. You will also need to set up IP forwarding, iptables etc. You also need to set option domain-name-servers to something reasonable. Since 127.0.0.1 appears in your list, I'm assuming that your box is running a DNS server, in which case probably best to set option domain-name-servers 192.168.0.x; as with option routers (note - no-one else can access your box as 127.0.0.1! :-) This is the contents of /etc/dhcpd.conf.etherboot.clients: host jennifer { hardware ethernet 00:80:AD:77:A3:9B; fixed-address 192.168.0.2; filenameboot-tulip.2.4.19-7mdk.nbi; } Looks resaonable. If you're using Etherboot 5.0.7 or greater, then the declaration is unnecessary; 5.0.7+ will automatically pick up the correct .nbi image file based on the PCI ID of the network card that they boot from. Michael
Re: [Cooker] Request for default LDAP server name in Drakx
On Sat, 3 Aug 2002, Buchan Milne wrote: Since we have LDAP and DHCP etc working, I normally do a network install, and DrakX (when given just my hostname) gets everything right. Setting up LDAP, it even pulls our prefix (dc=cae,dc=co,dc=za) out of the domainname it got from DHCP. The only irritation is that it defaults to localhost for the LDAP server. Somehow I don't see how I could manage to setup a working LDAP server on the machine by first boot so that I can log in ;-) Could this be changed to be something sensible like ldap.domainname (since it already has the correct domainname). I am quite sure that ldap.domainname will not have exist if it doesn't run ldap, and adding an A or CNAME record if the name doesn't exist is less effort than setting ldap.domainname on a number of machines, for each installation. This is the only change I have to make to get LDAP accounts working, so kudos on the rest. Would it not be better for LDAP to pull up an SRV record of the form _ldap._tcp.your.domain.nameIN SRV 0 0 389 your.ldap.server in order to locate an LDAP server for the domain? This would be consistent with the way that Kerberos locates its servers (LDAP and Kerberos in Win2k also use SRV records). Michael
[Cooker] samba-2.2.3a-10mdk (Mandrake 8.2 release) missing ACL support
Hi, samba-2.2.3a-10mdk was somehow built without ACL support, even though the SPEC file should include it. Proof: 1. rpm --requires samba does not list libacl as a requirement. 2. ldd /usr/sbin/smbd does not list libacl as a linked library 3. I cannot get Samba to display or modify ACLs created with setfacl on an XFS filesystem. 4. Rebuilding samba-2.2.3a-10mdk gives an smbd that does link against libacl (according to ldd). This is cc'ed to QA since there is nothing wrong with the source RPM file (no missing BuildRequires for libacl-devel etc.) and yet somehow the package managed to get built and distributed without ACL support. Michael Brown
[Cooker] Re: samba-2.2.3a-10mdk (Mandrake 8.2 release) missing ACL support
On Fri, 2 Aug 2002, Buchan Milne wrote: | samba-2.2.3a-10mdk was somehow built without ACL support, even though the | SPEC file should include it. | Proof: | 1. rpm --requires samba does not list libacl as a requirement. Seems to be so ... | 2. ldd /usr/sbin/smbd does not list libacl as a linked library Which is the reason for 1) | 3. I cannot get Samba to display or modify ACLs created with setfacl on | an XFS filesystem. This may not solve the actual problem, but it should help you: There are updated RPMs of 2.2.5a for 8.2 (and 8.1 and 8.0, but no acl support on 8.0 by default) with working acl support on ftp.samba.org and ~ other versions both on my site (http://ranger.dnsalias.com/mandrake/samba) and Sylvestre's (http://people.mandrakesoft.com/~staburet/samba) Thanks for that. | 4. Rebuilding samba-2.2.3a-10mdk gives an smbd that does link against | libacl (according to ldd). Could you try rebuilding the SRPM without libattr-devel? (I don't have an 8.2 box handy to try on, I could try on cooker ...) I don't have libattr-devel installed and it still builds OK: [fensys@koala bin]$ rpm -qa | egrep 'attr|acl' libacl1-2.0.0-1mdk libattr1-2.0.1-1mdk acl-2.0.0-1mdk libacl1-devel-2.0.0-1mdk [fensys@koala bin]$ pwd /home/fensys/rpm/BUILD/samba-2.2.3a/source/bin [fensys@koala bin]$ ldd smbd libacl.so.1 = /lib/libacl.so.1 (0x4001a000) libcups.so.2 = /usr/lib/libcups.so.2 (0x4002) libdl.so.2 = /lib/libdl.so.2 (0x4003b000) libnsl.so.1 = /lib/libnsl.so.1 (0x4003f000) libpam.so.0 = /lib/libpam.so.0 (0x40055000) libc.so.6 = /lib/libc.so.6 (0x4005d000) libattr.so.1 = /lib/libattr.so.1 (0x40199000) libssl.so.0 = /usr/lib/libssl.so.0 (0x4019c000) libcrypto.so.0 = /usr/lib/libcrypto.so.0 (0x401c9000) /lib/ld-linux.so.2 = /lib/ld-linux.so.2 (0x4000) so libattr-devel is not required to get smbd linked against libacl. Also, I can't remove any of the attr or acl packages without also removing libacl1-devel, which will break the already-existing BuildRequires in samba.spec: [root@koala root]# rpm -e libattr1 error: removing these packages would break dependencies: libattr.so.1 is needed by libacl1-2.0.0-1mdk libattr.so.1 is needed by acl-2.0.0-1mdk [root@koala root]# rpm -e libattr1 libacl1 acl error: removing these packages would break dependencies: libacl1 = 2.0.0 is needed by libacl1-devel-2.0.0-1mdk [root@koala root]# rpm -e libattr1 libacl1 acl libacl1-devel [fensys@koala SPECS]$ rpm -bp samba.spec error: failed build dependencies: libacl-devel is needed by samba-2.2.3a-10mdk so I really don't have any idea what the problem could be. If you can think of other things to try, I'll willingly do test builds on a clean 8.2 system. | This is cc'ed to QA since there is nothing wrong with the source RPM file | (no missing BuildRequires for libacl-devel etc.) and yet somehow the | package managed to get built and distributed without ACL support. I think we will have to take a look at this, since it recently happened also with 2.2.5a and acl 2.0.9. Geoffrey Lee reckoned it was due to the samba configure script not checking for libattr. And the samba rpm doesn't buildrequires libattr-devel, and the libacl-devel (which samba does buildrequire) doesn't require libattr. But I am quite sure I have built samba RPMs with acl support without libattr-devel. You are correct. It does link without libattr-devel: [bgmilne@bgmilne source]$ rpm -q libattr1-devel package libattr1-devel is not installed [bgmilne@bgmilne source]$ rpm -q libacl1-devel libacl1-devel-2.0.11-1mdk [bgmilne@bgmilne source]$ rm -f bin/smbd [bgmilne@bgmilne source]$ make /dev/null [bgmilne@bgmilne source]$ ldd bin/smbd |grep acl ~libacl.so.1 = /lib/libacl.so.1 (0x40024000) But also against libattr [bgmilne@bgmilne source]$ ldd bin/smbd |grep attr ~libattr.so.1 = /lib/libattr.so.1 (0x40288000) So, it seems like a packaging problem on libacl, libacl-devel should buildrequires libattr-devel How does this cause the observed problem (smbd not being linked against libacl)? Forgive me if I'm being dense. It's actually been quite a mission getting acl support in samba (it's been broken so many times ...), but it is worth it. Definitely. Michael Brown
Re: [Cooker] LTSP Admin. Tool
On Thu, 6 Jun 2002, Stew Benedict wrote: If I produce an KDE-based tool for installation, configuration, and management of LTSP desktop servers, what would be my chances or what would be required to have it included in the Mandrake distrobution? Limiting it to be KDE-specific may not be a good idea, but that of course is left to those paying the money and doing the work to decide ;-) I will be happy to package this for contribs, as I believe Mandrake + LTSP is the answer for education in Africa (and business too!). Of a bigger concern is getting LTSP packages into Mandrake also (including kernels, X, scripts etc), but a graphical configuration tool would be great. openmosix +LTSP would be awesome (hint, hint ...) Working on that package now (LTSP). It's actually possible to reproduce the entire functionality of LTSP (complete with all local apps) in around 400k of packages. I've done so; you can find them at www.fensystems.co.uk/RPMS.fensys - download the mkinitrd-net terminal-server RPMs, install and enjoy. Only reason I haven't uploaded them to contribs is because I gave up on waiting for things uploaded to contribs to make it into the distribution. Michael
Re: [Cooker] LTSP Admin. Tool
On Fri, 7 Jun 2002, Buchan Milne wrote: Working on that package now (LTSP). It's actually possible to reproduce the entire functionality of LTSP (complete with all local apps) in around 400k of packages. I've done so; you can find them at www.fensystems.co.uk/RPMS.fensys - download the mkinitrd-net terminal-server RPMs, install and enjoy. Only reason I haven't uploaded them to contribs is because I gave up on waiting for things uploaded to contribs to make it into the distribution. That's no excuse!!! If they don't get magically added to contribs, mail the cooker list, and if the stuff is useful, someone will surely get it in. However, I guess Stew will take a look at this (or can I get cracking?). Of course, links to the SRPMs (http://www.fensystems.co.uk/SRPMS.fensys/) are usually more useful. Hey, I figured that anyone smart enough to know how to use an SRPM would be able to do the mental regexp required to find them :-) Help yourself to the SRPMS. Bug reports are very welcome, patches even more so. My only request is to please e-mail patches to me instead of just adding them to your own SRPM, so that I can integrate them into the CVS repository. Michael
Re: [Cooker] LTSP Admin. Tool
On Fri, 7 Jun 2002, Stew Benedict wrote: Working on that package now (LTSP). It's actually possible to reproduce the entire functionality of LTSP (complete with all local apps) in around 400k of packages. I've done so; you can find them at www.fensystems.co.uk/RPMS.fensys - download the mkinitrd-net terminal-server RPMs, install and enjoy. Only reason I haven't uploaded them to contribs is because I gave up on waiting for things uploaded to contribs to make it into the distribution. That's no excuse!!! If they don't get magically added to contribs, mail the cooker list, and if the stuff is useful, someone will surely get it in. However, I guess Stew will take a look at this (or can I get cracking?). Of course, links to the SRPMs (http://www.fensystems.co.uk/SRPMS.fensys/) are usually more useful. I'll have a look. Sounds like much the same approach I was taking. It's a fairly neat way to solve the problem. Rather than creating a separate LTSP root, you just export your own root filesystem as read-only and all_squash (i.e. only files accessible to all users are visible on the export). You need to be running in security level 2 for this to work (in 3 and higher mode o+r is missing from many crucial files). This means that any application installed on the server is instantly available to be run as a local app on the diskless terminal. It also uses the stock Mandrake kernel, instead of requiring LTSP kernels. A small patch to Etherboot (which has been added to Etherboot CVS; I have commit access) causes the NIC to identify itself to the DHCP server. The mkinitrd-net package takes care of creating initrds for each network module and sets up the DHCP server so that the correct kernel+initrd automatically gets selected by the Etherboot code. Very painless. Michael
Re: [Cooker] LTSP Admin. Tool
On Fri, 7 Jun 2002, Stew Benedict wrote: Of course, links to the SRPMs (http://www.fensystems.co.uk/SRPMS.fensys/) are usually more useful. I'll have a look. Sounds like much the same approach I was taking. It's a fairly neat way to solve the problem. Rather than creating a separate LTSP root, you just export your own root filesystem as read-only and all_squash (i.e. only files accessible to all users are visible on the export). You need to be running in security level 2 for this to work (in 3 and higher mode o+r is missing from many crucial files). This means that any application installed on the server is instantly available to be run as a local app on the diskless terminal. The only issue I see immediately with that is if you're hosting clients that are other archs. I was building a seperate ltsp root, from the live system, which also has the same problem. I suppose in practice, that scenario probably doesn't arise that often. Probably not often enough to justify the (rather large) extra effort. FHS does sort of allow for one root filesystem supporting multiple architectures, doesn't it? Should it not be possible to have e.g. a /lib64 directory alongside /lib? Maybe RPM could be extended to allow installation of foreign architecture packages in this way? That would be a more generic solution. A small patch to Etherboot (which has been added to Etherboot CVS; I have commit access) causes the NIC to identify itself to the DHCP server. The mkinitrd-net package takes care of creating initrds for each network module and sets up the DHCP server so that the correct kernel+initrd automatically gets selected by the Etherboot code. Very painless. Sound like you've got it all worked out. Wish I knew this a week ago :) Join the Etherboot-developers list; I did announce it there :-) Michael
Re: [Cooker] LTSP Admin. Tool
On Fri, 7 Jun 2002, Matthew Tedder wrote: It's a fairly neat way to solve the problem. Rather than creating a separate LTSP root, you just export your own root filesystem as read-only and all_squash (i.e. only files accessible to all users are visible on the export). You need to be running in security level 2 for this to work (in 3 and higher mode o+r is missing from many crucial files). Seems a bit inflexible and a method prone to many unknowns. That would make me nervous until and unless I saw it running for a time at many a variety of sites to ensure it can still be twisted to meet the specific requirements of each. Not at all inflexible if you use ClusterNFS or something similar, since it allows you to create files that appear differently for each client if so desired. This means that any application installed on the server is instantly available to be run as a local app on the diskless terminal. Yes... I very nice upside.although there are good arguments for not having this...including security, differentiation for different hardware, etc. What's the security argument? Differentiation for different hardware is a non-issue: see below. It also uses the stock Mandrake kernel, instead of requiring LTSP kernels. A small patch to Etherboot (which has been added to Etherboot CVS; I have commit access) causes the NIC to identify itself to the DHCP server. The mkinitrd-net package takes care of creating initrds for each network module and sets up the DHCP server so that the correct kernel+initrd automatically gets selected by the Etherboot code. Very painless. I was under the impression that a lot of work was done in getting the images for LTSP as it is. Each different piece of hardware may very likely require different kernals. Not only that, but different X servers. The ThinkNIC, for example, requires the xvesa server in order not to blur graphics on screen. Many particularities exist for various different motherboards. Intel's DX810E requires USB drivers for the mouse/keyboard and printer. This is all likely to be very different then for the host. Every piece of hardware doesn't require a different kernel: that's what kernel loadable modules are for. X servers and some other hardware will require slightly different *configuration files* for some clients, yes. There are many ways to solve this problem, for example: 1. XFree86: have X started up using /etc/X11/XF86Config.hostname, falling back to /etc/X11/XF86Config if this doesn't exist (trivial). If you want to be more sophisticated, use a PCI scan for the graphics card and use /etc/X11/XF86Config.PCIID. 2. Any software package: use ClusterNFS' capability for serving up different files to different clients. If you create the file /etc/X11/XF86Config$$CLIENT$$ then any NFS client will see the contents of this file in place of /etc/X11/XF86Config. This approach (single software installation, differentiation of the configuration files) is far more elegant than duplicating (triplicating, etc.) package installations. If you take a look at the terminal-server package, you'll see the way that X configuration is handled: /etc/X11/XF86Config$$CLIENT$$ is symlinked to /tmp/XF86Config.test. /tmp is mounted as tmpfs on the clients (as indicated via /etc/fstab$$CLIENT$$ - obviously the clients need a different fstab to the server). This means that you can boot up a client, with the root filesystem mounted read-only via NFS, and immediately run XFdrake to generate your X config files in /tmp (XFdrake is available to run on the client simply because it is installed on the server). scp these back to the server (again, scp is installed on the server and therefore available on the client) and away you go. I have a planned improvement: X should fall back to framebuffer mode if it hasn't been configured yet for that machine (in the same way that the graphical installer uses framebuffer X). The share your own root filesystem approach is an order of magnitude easier to maintain than the separate ltsp root approach. If you install mkinitrd-net and terminal-server then you have a working terminal server, with the ability to boot from any network card you have a module for and to run any software that you have installed on the server. Nothing else needs to be done - just install and play. Michael
Re: [Cooker] Mounting /usr read-only
On Thu, 30 May 2002, Philippe Coulonges wrote: I would like to have /usr mounted as read-only. To achieve that, I have to ajust some things in the distro at every update. Could it be fixed, has I think variable files does not have their place there. This include kscd's /usr/share/apps/kscd/cddb/... Maelstrom and some other games scores files. It's actually possible to have almost the whole filesystem (from / downwards) mounted read-only or using tmpfs. This means that your box is pretty much guaranteed to withstand a kernel panic or power failure without any ill-effects on the filesystem. You have to patch initscripts a little to prevent complaints at startup. I did submit the patches a while back but they were rejected, because no-one could see why having / mounted ro would be useful... Michael
[Cooker] kdmrc AllowShutdown security bug [severity: critical]
In /usr/share/config/kdm/kdmrc: the AllowShutdown entry appears in the sections [X-*-Greeter] AllowShutdown=Root ... [X-:*-Greeter] AllowShutdown=All instead of [X-*-Core] AllowShutdown=Root ... [X-:*-Core] AllowShutdown=All The result is that kdm assumes the entry is missing and defaults to [X-*-Core] AllowShutdown=All ... [X-:*-Core] AllowShutdown=All thereby allowing any user on a *remote* X display to shut down or reboot the machine without having to supply the root password. This bug is present in kdebase-3.0.1-10mdk.i586.rpm (latest Cooker release). [ I tried to use Bugzilla to report this, but the Bugzilla system is unusable for many different reasons. ] Michael Brown http://www.fensystems.co.uk
Re: [Cooker] (HELP!) (rpm) (libtool) (going mad)
On Fri, 31 Aug 2001, Grégoire Colbert wrote: I never heard before about libtool, so I'm quite puzzled. I used the following lines in my spec file : %prep rm -rf %{buildroot} rm -rf $RPM_BUILD_DIR/%name-%version %setup %build ./configure --prefix=%_prefix %make prefix=%{buildroot}%_prefix %install %make prefix=%{buildroot}%_prefix install Did you try using: %prep %setup %build %configure %make %install rm -rf $RPM_BUILD_ROOT %makeinstall These generic instructions should work for most small packages that have configure scripts, including (I think) ones that use libtool. Certainly I remember seeing references to libtool flash past during the %build section and I have ended up with working packages. HTH, Michael
Re: [Cooker] Rpm group for new content filtering package
On Wed, 29 Aug 2001, Grégoire Colbert wrote: I'm doing a rpm for a content filtering web proxy, what would be a appropriate rpm group for that? My guess is: System Environment/Daemons Please repond swiftly See: http://www.linux-mandrake.com/en/howtos/mdk-rpm/mdk-groups.html I would say that Networking/WWW could be fine since System Environment/Daemons is not a Mandrake group (as long as the above document is up to date, obviously). FWIW, Squid (a web proxy) and SquidGuard (a content filter) are both in the group System/Servers. Michael
Re: [Cooker] One more... YAIR (21 aug 2001)
On Wed, 22 Aug 2001, Claudio wrote: LAN browsing in Konqueror does not work. Anyone had it working fine? Had it working perfectly. Download and install the package lisa-browselan from http://www.fensystems.co.uk/lisa-browselan-0.1-2mdk.noarch.rpm (source at http://www.fensystems.co.uk/lisa-browselan-0.1-2mdk.src.rpm) and see if it works. You may need to issue a service lisa start after installing. Since this is around the 5th request for it, I've re-uploaded it to /incoming. Maybe Lenny will notice it this time ;-) Shouldn't that package be installed by default with KDE than? It's like a small but very important plugin. The actual program (/bin/lisa) is part of the KDE packages. The lisa-browselan package provides an init.d script and the configuration files to run it. Since it listens on port 7741 it probably counts as a server and therefore is something you might not want to install and run just because you want KDE? I don't know - what does anyone else think? Michael
[Cooker] Segfault during installer - report.bug available
Trying installation with DrakX V1.544. Graphical mode locks up on language selection as previously reported. Tried a text install; failed during package selection with a segfault: Seems like memory is missing as install crashes. report.bug is available from http://www.fensystems.co.uk/report.bug to reduce list traffic. Michael
Re: [Cooker] One more... YAIR (21 aug 2001)
On Tue, 21 Aug 2001, Claudio wrote: Just performed a clean installation of today's cooker. Here you are a short report of what I found: NEGATIVE THINGS: snip 8) LAN browsing in Konqueror does not work. Anyone had it working fine? Had it working perfectly. Download and install the package lisa-browselan from http://www.fensystems.co.uk/lisa-browselan-0.1-2mdk.noarch.rpm (source at http://www.fensystems.co.uk/lisa-browselan-0.1-2mdk.src.rpm) and see if it works. You may need to issue a service lisa start after installing. Since this is around the 5th request for it, I've re-uploaded it to /incoming. Maybe Lenny will notice it this time ;-) Michael
[Cooker] [Beta1] DrakX V1.541 critical failure
Attempted cooker installation with DrakX V1.541 on an unbranded laptop (it's an experimental machine that I have on loan). Installation via NFS. PCMCIA NIC detection, DHCP, stage 2 load etc. is all fine. Immediately after GTK portion starts, I get: GLib-WARNING **: getpwuid_r(): failed due to: No such user 0. Gdk-ERROR **: BadAccess serial 120 error_code 10 request_code 143 minor_code 1 Gdk-ERROR **: BadShmSeg serial 121 error_code 155 request_code 143 minor_code 5 install exited abnormally :-( sending termination signals...done etc., etc., etc. Interestingly, a quick glance at console F3 shows * warning: rm of /usr/share/locale_special/af/LC_MESSAGES/libDrakX.mo failed: Read-only file system * warning: mkdir: error creating directory /usr/share/locale_special/en/LC_MESSAGES: Read-only file system Why is DrakX attempting to write to the NFS-mounted /usr? The file (../af/...) exists and is the first file under the /locale_special tree The directory (.../en/...) does not exist under /locale_special/, it exists under /locate/. Cannot send a full bug report as shell on console F2 is killed when the install exits abnormally (which happens as soon as the error occurs). Michael
[Cooker] Re: [Beta1] DrakX V1.541 critical failure
On Sun, 19 Aug 2001, Pixel wrote: [...] Gdk-ERROR **: BadAccess serial 120 error_code 10 request_code 143 minor_code 1 Gdk-ERROR **: BadShmSeg serial 121 error_code 155 request_code 143 minor_code 5 the X server must be getting crazy. I don't know why... These error messages appear when I use a remote X server (display=...). If I use the local X server then the machine just appears to hang once the language selection dialog appears. When I say hang, I mean that it fails to respond to anything, even Ctrl-Alt-(Del | Bksp). I presume the same error is happening, but I can't get to the text console to see the error messages. Michael
[Cooker] gendistrib reports lots of packages requiring a rebuild
Cooker i586 and contribs synchronised from fr.rpmfind.net at 22:40 GMT on 19/8/01. Running ./misc/gendistrib --distrib . generates a whole host of errors of the form xdelta-2.0-0.Beta1.2mdk require [libdb-3.1.so] which is not available in any medium listed Requirements are: libdb-3.1.so HTML-Embperl libgdbm1 libgdbm1-devel libSDL-1.1.so.0 libSDL1.1 libcdaudio libXbae.so.4 and many, many others. It's not just obscure contribs packages that need a rebuild; some of the packages in this situation are apache-devel apache-source php gtk-doc sympa Michael
Re: [Cooker] KDE 2.2
On Sat, 18 Aug 2001, Steve Guiwits wrote: Has anyone gotten the following erros while trying to install (via rpm) kde 2.2? error: failed dependencies: libpcreposix.so.0 is needed by kdelibs-2.2-1mdk libpcre.so.0 is needed by kdelibs-2.2-1mdk Yes. Upgrade libpcre0 and all will be fine. Unfortunately, libpcre0 is in contribs / unsupported for 8.0... (it's been moved to the main distro for cooker / 8.1). HTH, Michael
Re: [Cooker] LAN browsing
On Mon, 13 Aug 2001, David Walluck wrote: Do you have lisa running? I suspect not - it isn't defined as a service in 8.0 and nor (AFAIK) in Cooker. I did upload a package that fixed this problem, lisa-browselan, to /incoming a long time ago, but it seems to have disappeared into the void, along with a few other packages I uploaded such as squidGuard-blacklists, rdesktop and ucblogo. I'm beginning to wonder if there's any point in contributing stuff to Mandrake any more. I tried to enable lisa simply in my rc.local, but when I type lan:// I get an error. BTW, it says lisa and relisa need to be SUID root, then again so does smbmount and smbumount for normal users. I'm personally not against this, but I think Mandrake is. It doesn't need to be SUID root - the way I set it up it ran from a /etc/rc.d/init.d script so the binary itself didn't need to be SUID. I feel you on contribs... I know Lenny does his work on existing packages, but I am not sure how easy it is to get new packages in there. It's the lack of feedback that gets me. It's only by looking in /incoming and checking contribs that I get to find out if my packages have been lost. If no-one wants the things I'm packaging*, then it would be courteous if someone told me. * hard to believe, since I've had private requests for things I've packaged that people think I forgot to upload 'cos they can't find them in contribs! Michael
Re: [Cooker] LAN browsing
On Mon, 13 Aug 2001, Brook Humphrey wrote: I tried to enable lisa simply in my rc.local, but when I type lan:// I get an error. BTW, it says lisa and relisa need to be SUID root, then again so does smbmount and smbumount for normal users. I'm personally not against this, but I think Mandrake is. It doesn't need to be SUID root - the way I set it up it ran from a /etc/rc.d/init.d script so the binary itself didn't need to be SUID. I feel you on contribs... I know Lenny does his work on existing packages, but I am not sure how easy it is to get new packages in there. It's the lack of feedback that gets me. It's only by looking in /incoming and checking contribs that I get to find out if my packages have been lost. If no-one wants the things I'm packaging*, then it would be courteous if someone told me. * hard to believe, since I've had private requests for things I've packaged that people think I forgot to upload 'cos they can't find them in contribs! I would be another one that would like to know were to get this as I have clients that would like this functionality. OK - I've resisted doing this so far because I would prefer packages to go into contribs, but for those who want it, lisa-browselan is available from http://www.fensystems.co.uk/lisa-browselan-0.1-2mdk.src.rpm http://www.fensystems.co.uk/lisa-browselan-0.1-2mdk.noarch.rpm It works for me and for my customers; I make no claim of universal correctness! Michael
Re: [Cooker] LAN browsing
On Sun, 12 Aug 2001, Jean-Claude Tual wrote: I am new on this list, so this issue is perhaps very well known. When I tried to browse my LAN under Konqueror, I got a message box indicating : impossible to connect to host localhost. I have two ethernet cards : one on dhcp for inernet on cable and one for the LAN. If someone has an idea, thanks, Do you have lisa running? I suspect not - it isn't defined as a service in 8.0 and nor (AFAIK) in Cooker. I did upload a package that fixed this problem, lisa-browselan, to /incoming a long time ago, but it seems to have disappeared into the void, along with a few other packages I uploaded such as squidGuard-blacklists, rdesktop and ucblogo. I'm beginning to wonder if there's any point in contributing stuff to Mandrake any more. Michael
[Cooker] DrakX: partially interactive auto-installs: patch attached
I've adapted the Replay code in DrakX to make it possible to have an auto-install with selected stages being interactive, either via GTK, Newt or (theoretically) stdio. This introduces two new options into the auto-install configuration file: $o = { 'interactive' = 'gtk|newt|stdio', 'interactiveSteps' = [ 'e.g. doPartitionDisks', 'e.g. configureX', 'etc.' ] } 'Old-style' interactivity specified via $graphical=1 and push @graphical_steps, ... will still work (the code will automatically create the new options 'interactive' and 'interactiveSteps'). [ Note: I had to rearrange some code in install2.pm, so that 'interactive' doesn't automatically get set to 'gtk'. ] Here are the three patches: --- install_steps_newt.pm.orig Sat Jul 14 15:48:33 2001 +++ install_steps_newt.pm Sat Jul 14 17:00:24 2001 @@ -22,6 +22,7 @@ my $banner = translate(__(Linux-Mandrake Installation %s)); my $l = first(Newt::GetScreenSize) - length($banner) - length($_[0]) + 1; Newt::DrawRootText(0, 0, sprintf($banner, ' ' x $l . $_[0])); +Newt::Refresh; } sub new($$) { --- install2.pm.origSat Jul 14 18:36:38 2001 +++ install2.pm Sat Jul 14 18:53:23 2001 @@ -452,12 +452,6 @@ modules::read_stage1_conf($_) foreach /tmp/conf.modules, /etc/modules.conf; modules::read_already_loaded(); -$o-{interactive} ||= 'gtk'; -if ($o-{interactive} eq gtk availableMemory 22 * 1024) { - log::l(switching to newt install cuz not enough memory); - $o-{interactive} = newt; -} - if ($::auto_install) { require install_steps_auto_install; @@ -471,9 +465,16 @@ } unless ($::auto_install) { $o-{interactive} ||= 'gtk'; - requireinstall_steps_$o-{interactive}.pm; } + +if ($o-{interactive} eq gtk availableMemory 22 * 1024) { + log::l(switching to newt install cuz not enough memory); + $o-{interactive} = newt; +} +require install_steps_$o-{interactive}.pm if $o-{interactive}; + + eval { $o = $::o = install_any::loadO($o, patch) } if $patch; eval { $o = $::o = install_any::loadO($o, $cfg) } if $cfg; @@ -510,7 +511,6 @@ my $o_; while (1) { - requireinstall_steps_$o-{interactive}.pm; $o_ = $::auto_install ? install_steps_auto_install-new($o) : $o-{interactive} eq stdio ? --- install_steps_auto_install.pm.orig Sat Jul 14 11:14:50 2001 +++ install_steps_auto_install.pm Sat Jul 14 18:40:06 2001 @@ -7,8 +7,6 @@ @ISA = qw(install_steps); -@graphical_steps = qw(enteringStep beforeInstallPackages installPackages); - use modules; @@ -22,19 +20,34 @@ sub new { my ($type, $o) = @_; -if ($graphical) { - require install_steps_gtk; - push @ISA, 'interactive_gtk'; - foreach my $f (@graphical_steps) { +# Handle legacy options +$o-{interactive} ||= ( $graphical ? 'gtk' : undef ); +$o-{interactiveSteps} ||= [@graphical_steps]; + +if ($o-{interactive}) { +my $interactiveClass = install_steps_$o-{interactive}; + require $interactiveClass.pm; + push @ISA, interactive_$o-{interactive}; + push @{$o-{interactiveSteps}}, qw(enteringStep formatMountPartitions +beforeInstallPackages installPackages); + + *{install_steps_auto_install::wait_message} = sub { + local @ISA = ($interactiveClass, @ISA); + {'interactive::wait_message'}; + }; + + foreach my $f (@{$o-{interactiveSteps}}) { no strict 'refs'; - my $pkg = $install_steps_gtk::{$f} ? 'install_steps_gtk' : 'install_steps_interactive'; - log::l(install_steps_auto_install: adding function , $pkg, ::, $f); - *{install_steps_auto_install::$f} = sub { - local @ISA = ('install_steps_gtk', @ISA); + my $pkg = eval(\$${interactiveClass}::{$f}) ? $interactiveClass : + ( $install_steps_interactive::{$f} ? 'install_steps_interactive' : +undef ); +if ( $pkg ) { + log::l(install_steps_auto_install: adding function , $pkg, ::, $f); + *{install_steps_auto_install::$f} = sub { + local @ISA = ($interactiveClass, @ISA); {$pkg . '::' . $f}; - }; + }; + } } - goto install_steps_gtk::new; + goto {$interactiveClass.'::new'}; } else { (bless {}, ref $type || $type)-SUPER::new($o); }
Re: [Cooker] Cooker Compile
On Sat, 7 Jul 2001, Stefan van der Eijk wrote: But, the main problem is that the BuildRequires aren't accurate at the moment in cooker. Without the correct BuilRequires you're going to encounter many compile problems. I'm working on fixing them at the moment, but There are still +/- 380 packages that aren't compiling How are you detecting problems caused by 'non-fatal' missing BuildRequires? As an example: the recent missing requirement for libbzip2_1-devel and libtiff3-devel in kdelibs. If these are not installed then the package will be built without support for the relevant components, but no errors will be flagged. Is there any way to automatically detect this sort of omission? Michael
[Cooker] Bigmem/Highmem support in kernel and other kernel-enterprise issues
Am I correct in thinking that only the enterprise kernel supports physical memory 1GB? I'm a bit confused; I found this changelog entry: * Tue Mar 20 2001 Chmouel Boudjnah [EMAIL PROTECTED] 2.4.2-19mdk - Active Bigmem by default. but my current running kernel set (2.4.3-20mdk, as shipped with 8.0) shows that the configurations are config-2.4.3-20mdk:CONFIG_NOHIGHMEM=y config-2.4.3-20mdk:# CONFIG_HIGHMEM4G is not set config-2.4.3-20mdk:# CONFIG_HIGHMEM64G is not set config-2.4.3-20mdkenterprise:# CONFIG_NOHIGHMEM is not set config-2.4.3-20mdkenterprise:# CONFIG_HIGHMEM4G is not set config-2.4.3-20mdkenterprise:CONFIG_HIGHMEM64G=y config-2.4.3-20mdkenterprise:CONFIG_HIGHMEM=y config-2.4.3-20mdksmp:CONFIG_NOHIGHMEM=y config-2.4.3-20mdksmp:# CONFIG_HIGHMEM4G is not set config-2.4.3-20mdksmp:# CONFIG_HIGHMEM64G is not set i.e. Highmem (aka bigmem) disabled by default and enabled only for the enterprise kernel. What is the downside of enabling highmem support? I assume it breaks something, otherwise it would be enabled in all kernels. Another quick question: why is CONFIG_REISERFS_CHECK=y in the enterprise kernel (but not in the other kernels)? From the ReiserFS README: Real users, as opposed to folks who want to hack and then understand what went wrong, will want REISERFS_CHECK off. TIA, Michael
[Cooker] vnc-server-3.3.3r1-2mdk: Bug in %post
[mcb30@dolphin mcb30]$ rpm -q --scripts vnc-server postinstall script (through /bin/sh): if [ $1 = 1 ]; then /sbin/chkconfig --add vncserver; else if [ -f /var/lock/subsys/vncserver ]; then service vncserver restart /dev/null 2/dev/null || : fi fi # Install Doc install -d $RPM_BUILD_ROOT/vnc_docs ^^^ Shouldn't this be somewhere else?! This currently creates an empty vnc_docs directory in / Michael
Re: [Cooker] hdlist.cz?
On Fri, 29 Jun 2001, Jason Straight wrote: Is there documentation anywhere on creating the hdlist's? I'd like to keep my own local mirror of some updates and such for many machines and need to be able to generate my own hdlist.cz file from a directory full of RPMS. Thanks in advance. Basically, you just create a hdlists file which contains pointers to each of the RPMS directories (use the existing hdlists file as a template), then re-run gendistrib as $distroot/misc/gendistrib --distrib $distroot (where $distroot is the folder containing the root of your installation tree - e.g. distroot=/vault/mandrake/Mandrake/8.0/i586 in my case). This is assuming that you have a full local mirror of the standard 8.0/Cooker tree and wish to add one or more 'extra' RPM sets (I add Contribs and a few self-packaged, non-GPL extras like StarOffice in this way). HTH, Michael
Re: [Cooker] Installation can be easier
On Fri, 29 Jun 2001, unknown unknown wrote: Lucian Also, what do you all think about drag and drop installs, you have a CD in the drive, you are in gnome or nautilus, and you drag the game icon from the CD to a folder on your machine and it installs there. Possibly one of the worst ideas I've heard in recent times. RPM and the FHS are wonderful tools for keeping track of everything on your system. What you propose sounds like the Windows habit of installing everything in Program Files. You can already have click and play installs - just click on an RPM file and the installation process starts (IIRC - I never use a GUI for installation). Your proposal is a step several miles backwards from the current position. Automatic Installs would be hard to program i know that. But Image being able to just put a CD into your machine, and not having to click anything just watching whatever is on the CD load up on your screen. Just imagine the potential for viruses and other malignant entities. And the potential for destroying a working system by automatically adding a poorly-constructed package from something like a magazine cover CD. RPM, urpmi and friends (including apt, etc.) are the real killer applications on Linux, far surpassing anything on offer in the M$ world. Try comparing RPM to InstallShield - they're not just in different leagues; they're playing different sports! Michael
RE: [Cooker] Low Ram FTP Install
On Wed, 27 Jun 2001, Don Head wrote: I do read what you write, I just feel there should be some way to FTP install on a low ram computer. I also understand that certain sacrifices occur during the install (This is what I would like to learn more about). I'm not really sure how the install works, but from pieces I've picked up here and there, and a little consideration of my own, I think I have a little bit of an idea as to why the install takes so much RAM. I'll try and explain what I've come up with so far, and if I'm way off, I'm sure someone will correct me. First off, you have the boot disk. It's 1.4MB, and that has the kernel, common hardware and other modules, and gets the whole thing started. Can't remember if that's compressed or not, we'll assume so. We'll double it. +1.4MB X 2 =2.8MB Then, you have the second stage installer, which is downloaded once the boot disk gets up and running. If you take a look at your CD-ROM/FTP site, you'll notice that mdkinst_stage2.bz2 is 9.6MB, compressed. Uncompress that and you're looking at a big chunk of change. I haven't tried it, but I'll go with 2x compression for everything here on. +9.6MB X 2 =22MB AFAIK: With mounted (NFS, CD-ROM, HD) installs, you don't have to 'download' stage 2; there is an uncompressed copy already on the installation media so you just use it in situ. This is another, very significant reason why NFS installs are 'better' than FTP installs. Michael
[Cooker] Bug in gencert.sh from openldap-2.0.11-3mdk
--- gencert.sh Sun Jun 24 13:40:40 2001 +++ gencert.sh.new Wed Jun 27 14:45:59 2001 @@ -77,7 +77,7 @@ COMMONNAME=`hostname` -if [ ! -n $LANG ] +if [ ! -n $COMMONNAME ] then COMMONNAME=www.openldap.org fi I haven't bothered uploading a new SRPM since it's such a small change. Michael
Re: [Cooker] kdelibs woes (was: Mandrake 8.0 updates kill Arts)
On Wed, 20 Jun 2001, Vincent Danen wrote: How many more times should I report this bug? :-( Exactly when did you report the bug? Is it possible the packages were built prior to your reporting the bug? Did you check the cooker packages? rpm -qip the kdelibs packages and see if it was built after you reported it. Reported Jun 10, at which time 2.1.2-3mdk was not available on any of the mirrors that I check (both rpmfind and proxad.net). The problem therefore seems to be in the mirroring: both rpmfind and proxad are (AFAIR) primary mirrors and so I usually treat them as being definitive; if a package is not present then I assume it doesn't exist. QA process is not a 24hr deal... =) It sometimes takes a little bit longer, and kdelibs is a big package that needs thorough testing (this isn't cooker we're talking about!). So the delay is not in mirroring, but in QA. We could do it faster and not as thorough, but then we'd be doing it over again many more times, I'm sure. OK, that makes sense, but how about releasing the packages to cooker as soon as they are built and then releasing as updates once QA is complete? That way, everyone knows when an update is being worked on and you also benefit from extra testing. Just a thought, Michael
[Cooker] DrakX: kernel-enterprise as default kernel?
Is there a way to specify that the enterprise kernel should be used instead of the SMP kernel from within an auto-install file? My default_packages currently contains kernel and kernel-enterprise but not kernel-smp. All three kernels get installed, and lilo.conf contains entries for only kernel and kernel-smp (with SMP as the default). Also, the symlinks in /boot for the enterprise kernel do not get created, although if I uninstall and reinstall the enterprise kernel they get created correctly. All kernel versions are 2.4.3-20mdk, DrakX v1.510. I've had a quick look through the DrakX code and it doesn't seem to be aware of the enterprise kernel in the way that it is of the SMP kernel (if it detects a SMP machine then it seems to grab kernel-smp even though kernel-enterprise also provides SMP capability). TIA, Michael
Re: [Cooker] kdelibs woes (was: Mandrake 8.0 updates kill Arts)
On Thu, 21 Jun 2001, Vincent Danen wrote: QA process is not a 24hr deal... =) It sometimes takes a little bit longer, and kdelibs is a big package that needs thorough testing (this isn't cooker we're talking about!). So the delay is not in mirroring, but in QA. We could do it faster and not as thorough, but then we'd be doing it over again many more times, I'm sure. OK, that makes sense, but how about releasing the packages to cooker as soon as they are built and then releasing as updates once QA is complete? That way, everyone knows when an update is being worked on and you also benefit from extra testing. I just checked, and kdelibs in cooker is 2.2-0.alpha2... I don't know how we can get kdelibs 2.1.2 into cooker without causing even more problems... =) Remember, cooker moves fast and usually has the latest and greatest... versions that we support in updates cannot always use cooker as a vehicle for testing. Into unsupported (instead of cooker), then into updates after QA passed? Failing this, maybe updates should be announced on the cooker ML when they enter QA, so that bug-hunters are aware that a new build exists. Michael
Re: [Cooker] DrakX: kernel-enterprise as default kernel?
On 21 Jun 2001, François Pons wrote: All kernel versions are 2.4.3-20mdk, DrakX v1.510. I've had a quick look through the DrakX code and it doesn't seem to be aware of the enterprise kernel in the way that it is of the SMP kernel (if it detects a SMP machine then it seems to grab kernel-smp even though kernel-enterprise also provides SMP capability). We will take care of that, currently it is not working. only support for special kernel has been done. Thanks. I assume this means that the DrakX support for kernel-enterprise is not working, rather than that kernel-enterprise is not working? Michael
Re: [Cooker] kdelibs woes (was: Mandrake 8.0 updates kill Arts)
On Thu, 21 Jun 2001, Vincent Danen wrote: I just checked, and kdelibs in cooker is 2.2-0.alpha2... I don't know how we can get kdelibs 2.1.2 into cooker without causing even more problems... =) Remember, cooker moves fast and usually has the latest and greatest... versions that we support in updates cannot always use cooker as a vehicle for testing. Into unsupported (instead of cooker), then into updates after QA passed? Failing this, maybe updates should be announced on the cooker ML when they enter QA, so that bug-hunters are aware that a new build exists. I don't see how this will help. Security/bugfix updates are not open packages or public packages until they are published. In many cases, non-public bugs (security) are being fixed... making this available to cooker will cause us serious problems with other vendors. If anyone is sincerely interested in helping to beta test updates that make their way into rpmdrake, then please send me a note and we'll see if you qualify for our small secteam. Cooker is not updates, and vice versa. I typically update cooker *after* the updates for this same reason (non-disclosure until vulnerabilities are public). I'm sure you can appreciate this. Sure, but there's still the issue that triggered this thread: an update was built on Jun 4 but not released immediately (understandably), a new bug+fix was reported on Jun 10 (and several subsequent occasions) and received absolutely no response. There was (AFAIK) no way for me or anyone else to know that anyone was doing anything about kdelibs2.1.2 - even a short message such as kdelibs 2.1.2-3mdk has entered QA and will be in updates soon with no other details would have helped because I would then have waited for the release of the new build to see if the issue had already been resolved. Michael
Re: [Cooker] kdelibs woes (was: Mandrake 8.0 updates kill Arts)
On Tue, 19 Jun 2001, Vincent Danen wrote: Easy to resolve for those who know what they're doing. Now, picture the newbie to Mandrake who got talked into it by one of us trying to figure out why their sound just won't work after they did exactly what they were taught and trusted only RPM's from Mandrake's update facilities. Unless this problem is fixed by an official Mandrake RPM through the standard update facility, we have failed the new users of the Linux community and yet again shown it to be aimed at the technically elitist bunch. Yes! I have been waiting for someone to raise that point. Now add to the list, a fix for the Konqueror man page display, and us new ML8 users will be back to the same level of functionality that came in the box. I see that kdelibs2.1.2 has been rebuilt, but *without* the two missing BuildRequires that I pointed out several times: BuildRequires: libbzip2_1-devel libtiff3-devel As a result, the new update (2.1.2-3mdk) is *still* missing the man page display functionality. How many more times should I report this bug? :-( Exactly when did you report the bug? Is it possible the packages were built prior to your reporting the bug? Did you check the cooker packages? rpm -qip the kdelibs packages and see if it was built after you reported it. Reported Jun 10, at which time 2.1.2-3mdk was not available on any of the mirrors that I check (both rpmfind and proxad.net). The problem therefore seems to be in the mirroring: both rpmfind and proxad are (AFAIR) primary mirrors and so I usually treat them as being definitive; if a package is not present then I assume it doesn't exist. Michael
Re: [Cooker] kdelibs woes (was: Mandrake 8.0 updates kill Arts)
On Mon, 18 Jun 2001, Timothy Wagner wrote: Easy to resolve for those who know what they're doing. Now, picture the newbie to Mandrake who got talked into it by one of us trying to figure out why their sound just won't work after they did exactly what they were taught and trusted only RPM's from Mandrake's update facilities. Unless this problem is fixed by an official Mandrake RPM through the standard update facility, we have failed the new users of the Linux community and yet again shown it to be aimed at the technically elitist bunch. Yes! I have been waiting for someone to raise that point. Now add to the list, a fix for the Konqueror man page display, and us new ML8 users will be back to the same level of functionality that came in the box. I see that kdelibs2.1.2 has been rebuilt, but *without* the two missing BuildRequires that I pointed out several times: BuildRequires: libbzip2_1-devel libtiff3-devel As a result, the new update (2.1.2-3mdk) is *still* missing the man page display functionality. How many more times should I report this bug? :-( Michael
Re: [Cooker] kdelibs woes (was: Mandrake 8.0 updates kill Arts)
On Mon, 18 Jun 2001, Timothy Wagner wrote: Easy to resolve for those who know what they're doing. Now, picture the newbie to Mandrake who got talked into it by one of us trying to figure out why their sound just won't work after they did exactly what they were taught and trusted only RPM's from Mandrake's update facilities. Unless this problem is fixed by an official Mandrake RPM through the standard update facility, we have failed the new users of the Linux community and yet again shown it to be aimed at the technically elitist bunch. Yes! I have been waiting for someone to raise that point. Now add to the list, a fix for the Konqueror man page display, and us new ML8 users will be back to the same level of functionality that came in the box. I see that kdelibs2.1.2 has been rebuilt, but *without* the two missing BuildRequires that I pointed out several times: BuildRequires: libbzip2_1-devel libtiff3-devel As a result, the new update (2.1.2-3mdk) is *still* missing the man page display functionality. How many more times should I report this bug? :-( Michael
[Cooker] kdelibs-devel-2.1.2-2mdk: rebuild still required
On Mon, 11 Jun 2001, Michael Brown wrote: kdelibs-2.1.2-2mdk already needs a rebuild with added BuildRequires: libbzip2_1-devel libtiff3-devel to fix the problem of the missing bzip2 filter (and hence no man page browser), as reported in an earlier mail. Any chance of a rebuild of kdelibs - even just into unsupported? Michael
Re: [Cooker] Re: Locale settings and kdeglobals
On Mon, 11 Jun 2001, Michael Brown wrote: Is the installer supposed to set the Charset, Country and Language entries under [Locale] in /usr/share/config/kdeglobals? it used to, but it doesn't anymore. It must somewhere in kde scripts now. Well, it thinks I'm in Africa for some reason, straight after a clean install with 'lang'='en_GB' in the auto-install file... Found a bug: The startkde script (kdebase-2.1.1-13mdk) extracts the locale settings from the LANG environment variable, when it should (AFAICT) use the LANGUAGE environment variable. LANG contains only the language code (e.g. en), but not the country code (GB in en_GB). LANGUAGE contains both. startkde incorrectly uses LANG and so never picks up a country code. Michael
[Cooker] Probably a stupid question, but...
...how do you stop QT from attempting to use anti-aliased fonts on a display that doesn't support the RENDER extension? xdpyinfo indicates that RENDER is not available. Starting any QT/KDE program gives the message 'Xlib: extension RENDER missing on display :0.0.' QT is attempting to use anti-aliasing, because the fonts appear looking strangely distorted (imagine using AA fonts but quantising the alpha channel to on/off only and you get the picture). Have tried exporting QT_XFT=0, QT_XFT=false and various other combinations, all without making any difference. Using qt 2.3.0-3mdk, XFree86-4.0.3-7mdk Michael
Re: [Cooker] Probably a stupid question, but...
On Wed, 13 Jun 2001, Arnd Bergmann wrote: You cannot turn off the attempt to access the RENDER extension without recompiling Qt. However, you don't need to (and it won't improve the fonts), because your real problem is something else. You were right. I was convinced it was AA that was causing the problem, but I've just had a look at all the installed font packages and noticed that only XFree86-100dpi-fonts and urw-fonts are installed. Installing XFree86-75dpi-fonts fixes the problem. This was on a clean install from an auto-install file that (I now see) didn't explicitly specify any font packages. 100dpi and urw are required by other packages, so they got installed anyway. Ah well, that's something to remember for the future. Thanks for the help, Michael
Re: [Cooker] Re: Locale settings and kdeglobals
On Mon, 11 Jun 2001, Alexander Skwar wrote: unifying the storage of Pine, KMail, evolution and whatever other MUAs I What do you mean by this? They all can use IMAP/POP3, can't they? Mostly, yes. Most of them also use mbox format for storing mail locally, and it is possible (to some extent, have not investigated fully) to get them to use the same set of folders. For example, by editing kmailrc you can have KMail and Pine sharing the same local folders, so that you could switch seamlessly between the two. My aim in doing this is to make it easier for newbies to choose an e-mail package. At the moment, you really have to select a package before you start using it seriously. If they all used the same folder set, then you would be able to switch packages even after you'd already filled your folders with mail. It may not be possible to get them to share local folders nicely, in which case I'll give up and have all folders (not just inboxes) stored on an IMAP server. Michael
[Cooker] kdelibs-devel-2.1.2-2mdk: includes references to /home/baudensRPM_BUILD_ROOT
See below. End result is that KDE library documentation is unusable by KDevelop (without hacking). kdelibs-2.1.2-2mdk already needs a rebuild with added BuildRequires: libbzip2_1-devel libtiff3-devel to fix the problem of the missing bzip2 filter (and hence no man page browser), as reported in an earlier mail. Michael [mcb30@dolphin mcb30]$ gunzip -c /usr/share/doc/HTML/en/kdoc-reference/* | grep -i baudens BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//dcop BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//interfaces BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//kab BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//kdecore BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//kdedbcore BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//kdedbui BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//kdejava BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//kdeui BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//kfile BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//khtml BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//kio BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//kjs BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//kparts BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//kspell BASE URL=/home/baudens/tmp/kdelibs-2.1.2-2mdk-root//usr/share/doc/HTML/en//libkmid [mcb30@dolphin mcb30]$ rpm -qf /usr/share/doc/HTML/en/kdoc-reference/* kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk kdelibs-devel-2.1.2-2mdk
Re: [Cooker] Kdelibs missing files (was: From NG - soud in KDE onMdk8 lost...)
On Sat, 9 Jun 2001, Ural Khassanov wrote: The second problem is that updated kdelibs don't include bzip2 filter dll /usr/lib/kde2/kbzip2filter.* and /usr/share/services/kbzip2filter.desktop so updated konqueror stops viewing manpages. This is something I also found last night. The updated kdelibs is also missing kimg_g3.so and kimg_tiff.so. Interestingly, the PPC package with the exact same version number (kdelibs-2.1.2-2mdk) *does* include these. I am going to try a rebuild to see if I can find the problem today, because I need to demo one of these systems tomorrow and it would be good if the documentation was visible! Michael
Re: [Cooker] Auto-install and mount options
On Fri, 8 Jun 2001, Michael Brown wrote: i've just send a new version to our webmaster. In the meantime use http://perso.mandrakesoft.com/~prigaux/auto_inst.html written by [EMAIL PROTECTED] based on my previous doc. comments are welcome. (i've not finished commenting it) Thanks, will look soon and give feedback. Looks good, but would be improved by having some links within the document (e.g. a table of contents like the old one has). Also, here are a couple of contributions which may be of interest: I always change the first line of auto-install scripts to #!/usr/bin/perl -cw and make the scripts executable. After editing you can then simply execute the script to do a syntax check. It only saves three seconds, but it's three seconds that can occur hundreds of times. For a completely unattended install, you can add the following: 'autoExitInstall' = 1, 'postInstallNonRooted' = 'EOF', modprobe vfat EOF 'postInstall' = 'EOF', /bin/mkdir -p /mnt/floppy /bin/mount -t vfat /dev/fd0 /mnt/floppy /bin/dd if=/dev/`perl -ne 'print if s{/dev/(\D+)\d* /boot .*}{$1}' /etc/fstab` of=/mnt/floppy/pass.bs bs=512 count=1 sync umount /mnt/floppy sync EOF and to syslinux.cfg: default pass ... label pass kernel pass.bs At the end of the installation, this will write the boot sector from the hard disk to a file pass.bs on the installation floppy disk. The computer will then reboot, boot from the installation floppy and, if left unattended, will then boot from pass.bs which will boot the installed system. You therefore don't need to do anything once you've started off the auto-install : just have a cup of coffee and come back ten minutes later to find a system that's fully installed and also up and running. Michael
[Cooker] [Contrib-RPM] Uploaded to /incoming: lisa-browselan-0.1-2mdk
Name: lisa-browselan Relocations: (not relocateable) Version : 0.1 Vendor: MandrakeSoft Release : 2mdk Build Date: Thu 07 Jun 2001 23:20:23 BST Install date: (not installed) Build Host: dolphin.home Group : Networking/Other Source RPM: (none) Size: 3060 License: GPL Packager: Michael Brown [EMAIL PROTECTED] Summary : Lan Information Server (LISa) LAN browsing Description : LISa - the LAN Information Server - allows you to browse machines connected to your LAN in a way similar to Network Neighbourhood in the MS world. LISa allows you to browse using multiple protocols: SMB, FTP, HTTP and (allegedly) NFS. LISa has a flexible method of searching for hosts on the network. This package includes a default configuration that will browse for all hosts that offer SMB as one of their services. This will therefore pick up all Windows machines plus any boxes running Samba. To see LISa's browsing capabilities, open up Konqueror and point it at lan:/ * Thu Jun 07 2001 Michael Brown [EMAIL PROTECTED] 0.1-2mdk - Added /usr/share/config/kio_lanrc to avoid need to individually configure LAN browsing * Sat Jun 02 2001 Michael Brown [EMAIL PROTECTED] 0.1-1mdk - First Mandrake version
[Cooker] htdig binaries located in /usr/sbin? (affects kdevelop)
Is there any reason why the ht://Dig binaries (htsearch etc.) are located in /usr/sbin rather than /usr/bin? This puts them outside the search path of normal users (under any of the preset security levels) and so programs that try to use them (e.g. KDevelop) cannot find them without help. I guess that the htdig package should be modified, unless there is a reason why the binaries should be in /usr/sbin - Chmouel? Michael
Re: [Cooker] Kdelibs missing files (was: From NG - soud in KDE onMdk8 lost...)
On Sun, 10 Jun 2001, Michael Brown wrote: The second problem is that updated kdelibs don't include bzip2 filter dll /usr/lib/kde2/kbzip2filter.* and /usr/share/services/kbzip2filter.desktop so updated konqueror stops viewing manpages. This is something I also found last night. The updated kdelibs is also missing kimg_g3.so and kimg_tiff.so. Interestingly, the PPC package with the exact same version number (kdelibs-2.1.2-2mdk) *does* include these. I am going to try a rebuild to see if I can find the problem today, because I need to demo one of these systems tomorrow and it would be good if the documentation was visible! Found the problem: missing BuildRequires! kdelibs: BuildRequires: libbzip2_1-devel libtiff3-devel There may be other missing BuildRequires; I haven't looked in detail. A new kdelibs should probably be released via updates after rebuilding with these BuildRequires present, mainly because the missing bzip2 bits result in Konqueror silently failing to display any man pages. Michael
Re: [Cooker] Auto-install and mount options
On Sun, 10 Jun 2001, David Eastcott wrote: Also, here are a couple of contributions which may be of interest: I always change the first line of auto-install scripts to #!/usr/bin/perl -cw This would make a nice touch to the generated file. and make the scripts executable. After editing you can then simply execute the script to do a syntax check. It only saves three seconds, but it's three seconds that can occur hundreds of times. This might be better left to the discretion of the user, but perhaps adding a comment to go along with the above #!... would let the user know he has the option... --- usr/bin/perl-install/install_any.pm Wed Apr 18 09:19:12 2001 +++ usr/bin/perl-install/install_any.pm.new Sun Jun 10 16:57:12 2001 @@ -629,7 +629,11 @@ require Data::Dumper; join('', - # You should always check the syntax with 'perl -cw auto_inst.cfg.pl' before testing\n, + #!/usr/bin/perl -cw\n, + #\n, + # You should check the syntax of this file before using it in an +auto-install. You\n, + # can do this with 'perl -cw auto_inst.cfg.pl' or by executing this file +(note the\n, + # '#!/usr/bin/perl -cw' on the first line).\n, Data::Dumper-Dump([$o], ['$o']), if_($replay, qq(\npackage install_steps_auto_install;), q( $graphical = 1; You therefore don't need to do anything once you've started off the auto-install : just have a cup of coffee and come back ten minutes later to find a system that's fully installed and also up and running. Neat approach. I gather that you left the orginal section declarations in syslinux.cfg and just added a new one? So then you would boot up with the floppy, and select 'linux' on the command line to allow the initial install to proceed? Yes (although in my case I also replaced the boot message and added options for server and client which use my custom auto-install files). Would it not be easier to just wait a few seconds more till the 2nd stage install started, then pop the floppy out to acheive the same result? This would ensure that the computer really does boot correctly from disk. I often want to start an auto-install and then get on with something else, often in a different location. The pass.bs technique minimises the time I have to wait by the machine I'm installing and, once it's written, is transparent and easy to use (you don't have to do anything!). Michael
[Cooker] Locale settings and kdeglobals
Is the installer supposed to set the Charset, Country and Language entries under [Locale] in /usr/share/config/kdeglobals? Michael
[Cooker] Re: Locale settings and kdeglobals
On Mon, 11 Jun 2001, Pixel wrote: Is the installer supposed to set the Charset, Country and Language entries under [Locale] in /usr/share/config/kdeglobals? it used to, but it doesn't anymore. It must somewhere in kde scripts now. Well, it thinks I'm in Africa for some reason, straight after a clean install with 'lang'='en_GB' in the auto-install file... I don't have the energy to track it down right now, but if anyone knows where the relevant code is, please tell me! I'm currently busy fixing minor packaging bugs in the KDevelop et al. collection and after that my ToDo list includes packaging up the SquirrelMail webmail package and unifying the storage of Pine, KMail, evolution and whatever other MUAs I can find... Michael
[Cooker] Auto-install and mount options
Is there a way to specify mount options for partitions that are *created* during installation (i.e. those specified in the 'partitions' section rather than those specified in the 'manualFstab' section)? I have not been able to trace it all the way through the installer code to find out a definitive answer. Setting 'options' within a 'partitions' entry has no effect: 'partitions' = [ { 'mntpoint' = '/', 'size' = 523089, 'type' = 387, 'options' = 'rw,suid,dev,exec,auto,nouser,async' }, etc. and neither does adding a manualFstab entry for a mount point defined in 'partitions': 'partitions' = [ { 'mntpoint' = '/boot', 'size' = 19089, 'type' = 131 }, ... 'manualFstab' = [ { 'mntpoint' = '/boot', 'options' = 'ro,nosuid,nodev,noexec,auto,nouser,sync' }, g_auto_install doesn't help because there is (AFAICT) no way to set mount options from within the GUI. On a more general note, is there any comprehensive documentation on DrakX and the contents of the auto-install file. The 'kickstart' page (http://www.linux-mandrake.com/drakx/auto_inst.html) is still very incomplete and contains some entries like: 1.4.11. printer Various parameters for configuring your printer, being local, remote, remote SMB, remote NCP,... which don't actually document the options. Michael
Re: [Cooker] Auto-install and mount options
On Fri, 8 Jun 2001, Pixel wrote: i don't think there's any way to achieve what you want OK, I'll use perl -pi on /etc/fstab in the postInstall instead. 'partitions' = [ { 'mntpoint' = '/', 'size' = 523089, 'type' = 387, 'options' = 'rw,suid,dev,exec,auto,nouser,async' }, etc. i'll see if it's easy to add this. Thanks. and neither does adding a manualFstab entry for a mount point defined in 'partitions': [...] 'manualFstab' = [ { 'mntpoint' = '/boot', 'options' = 'ro,nosuid,nodev,noexec,auto,nouser,sync' }, did you try: manualFstab = [ { device = '/dev/hda1', options = 'whatever' } ], No, because there's no guarantee that partitions will be allocated as hdaX (e.g. I have some systems which have only hde and no hda). g_auto_install doesn't help because there is (AFAICT) no way to set mount options from within the GUI. On a more general note, is there any comprehensive documentation on DrakX and the contents of the auto-install file. The 'kickstart' page (http://www.linux-mandrake.com/drakx/auto_inst.html) is still very i've just send a new version to our webmaster. In the meantime use http://perso.mandrakesoft.com/~prigaux/auto_inst.html written by [EMAIL PROTECTED] based on my previous doc. comments are welcome. (i've not finished commenting it) Thanks, will look soon and give feedback. Michael
Re: [Cooker] various gnome apps hanging
On Wed, 6 Jun 2001, Frederic Crozat wrote: I complained a bit ago about galeon hanging. This seems to be spreading to other multithreaded gnome apps on my Cooker workstation. I am now seeing pan hang in the exact same way as galeon. In both cases, the program does not actually hang. It does however block in a select with no application gui refresh. Could you try to rename /etc/gtk/gtkrc to /etc/gtk/gtkrc.old and see if you still have freeze ? (Of course, you will be back to default gtk theme).. Dont' know about galeon, but it fixes the problems I've been having with Gimp segfaulting. Michael
Re: [Cooker] DrakX / LILO / MBR
On 6 Jun 2001, François Pons wrote: Forget it if you want; it's just an option that I came across a need for yesterday (for boring reasons that I won't go in to), and I thought I may as well contribute the patch that implements it, since it's such a simple change. We will add a onMbr flag and keep (as obsolete but for compability) crushMbr, define it to force onMbr property so that only one is defined... Sounds good to me. Michael
[Cooker] Improving robustness of startup/shutdown scripts
I have come across a few badly-behaved startup scripts (such as for vncserver) which can, under some circumstances, expect user input when called as /etc/rc.d/init.d/whatever start|stop. During system startup, stdout does not appear, so the user never sees the prompt - it just looks as though the machine has hung partway through starting up the service. A fix for this problem could be to patch the definition of action in /etc/rc.d/init.d/functions, so that the init.d scripts are always run with stdin redirected to /dev/null. Something like action() { ... initlog $INITLOG_ARGS -c $GPRINTF_REST /dev/null success ... ... Is this a reasonable idea? Michael
Re: [Cooker] DrakX / LILO / MBR
On Tue, 5 Jun 2001, Pixel wrote: Is there any way to tell DrakX's auto-install mode to install LILO onto /dev/hda1 instead of /dev/hda ? I've looked through the code and the nearest option I can find seems to be well, you could use bootloader = { boot = 'hda1' }, Not quite as elegant: using preserveMbr will cause it to automatically choose the partition containing /boot. preserveMbr will work even if there is no hda on the system (e.g. an Abit VP6 using only the HPT370 IDE channels). -my ($onmbr, $unsafe) = $lilo-{crushMbr} ? (1, 0) : suggest_onmbr($hds); +my ($onmbr, $unsafe) = $lilo-{preserveMbr} ? (0, 0) : ( $lilo-{crushMbr} ? (1, 0) : suggest_onmbr($hds) ); is that really needed? It's not needed any more than crushMbr is, no. It's just that at the moment there is an option to force the use of the MBR (crushMbr) without a corresponding option to force the non-use of the MBR (preserveMbr). (To anyone who hasn't been following this: setting crushMbr=0 does not force the non-use of the MBR, so preserveMbr is not a redundant option). Forget it if you want; it's just an option that I came across a need for yesterday (for boring reasons that I won't go in to), and I thought I may as well contribute the patch that implements it, since it's such a simple change. Michael
Re: [Cooker] DrakX / LILO / MBR
On 6 Jun 2001, François Pons wrote: Is there any way to tell DrakX's auto-install mode to install LILO onto /dev/hda1 instead of /dev/hda ? I've looked through the code and the nearest option I can find seems to be $o - {bootloader}{crushMbr} which, if set to 1, causes LILO to install onto /dev/hda but if set to 0 causes LILO to intelligently decide whether to install onto the MBR or the partition. How about the following patch: snip This adds an extra option $o - {bootloader}{preserveMbr} which, if set to 1, will force LILO to leave the MBR intact. This auto-install option must obviously be used with care because if the MBR is not valid to start with then the resulting system will not boot. That why crushMbr is used, as avoiding mbr install is not normal and is meaning a mbr signature is unknown to DrakX, so what are you using to boot linux in such case, and can you send us the mbr of your disk with : dd if=/dev/hda of=/tmp/mbr count=1 bs=512 because actually, it could means we expect linux unable to boot on the system as DrakX doesn't known anything about the installed mbr. You can install lilo to /dev/hda1 (or other partition) instead of /dev/hda and then a 'standard' DOS MBR will happily boot lilo if the lilo partition is marked active. (Tried and tested). At the moment, DrakX will always overwrite a DOS MBR. You might also want to leave the MBR intact if you were auto-installing a dual-boot machine and wanted to use the pre-existing boot loader as the top-level boot loader: e.g. if you wanted Linux to appear as an option on NT's boot menu by adding an entry to boot.ini. (Yes, I know that you'd have to add some code manually to the postInstall to write the correct entry into boot.ini). Michael
Re: [Cooker] [Contrib-RPM] cygwin-xfree86-20010525-1mdk
On 5 Jun 2001, Chmouel Boudjnah wrote: This package provides a Windows ZIP file, shared via Samba, that Windows clients can grab and install easily, without having to install Cygwin separately. Do NOT install this package if you already use Cygwin on your Windows boxes; it is likely that this package will overwrite some of your files (since Windows doesn't have a proper package management tool :-) we don't shipp applications for MS-Windows [tm][1] in our distrib... Footnotes: [1] even free software one.. OK, as long as you've noticed the important bit of the description: the bit that says Cygwin/XFree86 is an X server for Windows. It enables you to open up a graphical login session to your Linux-Mandrake box. I say this only because you didn't quote that part and I want to make sure that you realise what the package does (enables Windows users to try out Mandrake with minimal effort) before rejecting it. Michael
Re: [Cooker] DrakX / LILO / MBR
On Tue, 5 Jun 2001, Blue Lizard wrote: Is there any way to tell DrakX's auto-install mode to install LILO onto /dev/hda1 instead of /dev/hda ? I've looked through the code and the nearest option I can find seems to be $o - {bootloader}{crushMbr} which, if set to 1, causes LILO to install onto /dev/hda but if set to 0 causes LILO to intelligently decide whether to install onto the MBR or the partition. How about the following patch: snip This adds an extra option $o - {bootloader}{preserveMbr} which, if set to 1, will force LILO to leave the MBR intact. This auto-install option must obviously be used with care because if the MBR is not valid to start with then the resulting system will not boot. so wait...im slower than normal people...im special Did u check through the expert install reference to observe the function passed to on the bootloader install graphic signal? Then just recreate the call in your autoinstall mode. almost cutpaste but following whatever signal recognition scheme is used in the autoinstall code. Honestly that is the philosophy i would use but then im not quite familiar with any of the drakx code, just the guy that wishes for more features but doesnt get em. snip Sort of no longer auto but then you seem to be willing to modify code on a case by case basis anyway |-|:P The point is that I don't want to hack the install code for a specific case: I want to add an extra option so that the whole thing becomes configurable via the normal mechanism for auto-installs. I spent over an hour going through the code and this was the cleanest change I could find: it's a one-line patch, it adds one extra option and it won't affect anything else (previous auto-install files will continue to function as before). If you can find something cleaner, please do! Michael
[Cooker] DrakX / LILO / MBR
Is there any way to tell DrakX's auto-install mode to install LILO onto /dev/hda1 instead of /dev/hda ? I've looked through the code and the nearest option I can find seems to be $o - {bootloader}{crushMbr} which, if set to 1, causes LILO to install onto /dev/hda but if set to 0 causes LILO to intelligently decide whether to install onto the MBR or the partition. How about the following patch: --- usr/bin/perl-install/bootloader.pm.orig Tue Jun 5 15:42:30 2001 +++ usr/bin/perl-install/bootloader.pm Tue Jun 5 19:09:57 2001 @@ -222,7 +222,7 @@ require c; c::initSilo() if arch() =~ /sparc/; -my ($onmbr, $unsafe) = $lilo-{crushMbr} ? (1, 0) : suggest_onmbr($hds); +my ($onmbr, $unsafe) = $lilo-{preserveMbr} ? (0, 0) : ( $lilo-{crushMbr} ? (1, +0) : suggest_onmbr($hds) ); add2hash_($lilo, arch() =~ /sparc/ ? { default = linux, This adds an extra option $o - {bootloader}{preserveMbr} which, if set to 1, will force LILO to leave the MBR intact. This auto-install option must obviously be used with care because if the MBR is not valid to start with then the resulting system will not boot. Has been tested and does work. Michael
Re: [Cooker] /incoming is full again...
On 3 Jun 2001, andre wrote: Can I use the donations page to donate some money specifically to be used for buying an extra disk for /incoming? :-) was that after you send the X server? Not after, during. Michael
Re: [Cooker] [Contrib-RPM] cygwin-xfree86-20010525-1mdk
On Sun, 3 Jun 2001 [EMAIL PROTECTED] wrote: Did you made the zipfile yourself or did you find it somewhere. In that i´m interested. cos i want the zipfile The zipfile gets built from various .tar.gz/bz2s from the Cygwin site when you build the source RPM. The spec file also tweaks it a bit to add code to auto-detect the screen size and resolution and modifies the Cygwin startup batch file so that it starts the X server instead of trying to start bash (which is not included in this package, since it's not meant to be a full Cygwin installation). The X server code itself is taken unmodified from Cygwin. The packaging (zip file creation, extra tweaks) is mine. If you want the zip file, you can grab it from http://www.fensystems.co.uk/cygxf86/ or alternatively grab the SRPM, do an rpm -bi and copy the zip file out of the buildroot instead of building the binary package. Michael
[Cooker] LAN browsing (lisa) non-functional in 8.0
LAN browsing using LISa (which provides the lan:/ and rlan:/ kios) doesn't work (doesn't do anything). The binaries (/usr/bin/[res]lisa) are present but the lisarc file doesn't exist and there's no script in /etc/rc.d/init.d to start lisa at boot time. I am currently putting together a small package lisa-browselan that will contain the missing bits - expect an upload to /incoming later today. For the sake of completeness, the following should also be added to /etc/services - Chmouel? --- /etc/services Mon May 28 00:11:57 2001 +++ ./services.with-lisaSat Jun 2 13:49:02 2001 @@ -9600,6 +9600,7 @@ hostmon 5355/udp# hostmon uses TCP (nocol) ircd6667/tcp# Internet Relay Chat ircd6667/udp# Internet Relay Chat +lisa7741/tcp# Lan Information Server tproxy 8081/tcp# Transparent Proxy tproxy 8081/udp# Transparent Proxy mandelspawn 9359/udpmandelbrot # network mandelbrot
[Cooker] [Contrib-RPM] lisa-browselan-0.1-1mdk
Name: lisa-browselan Relocations: (not relocateable) Version : 0.1 Vendor: MandrakeSoft Release : 1mdk Build Date: Sat 02 Jun 2001 18:24:17 BST Install date: (not installed) Build Host: dolphin.home Group : Networking/Other Source RPM: (none) Size: 2737 License: GPL Packager: Michael Brown [EMAIL PROTECTED] Summary : Lan Information Server (LISa) LAN browsing Description : LISa - the LAN Information Server - allows you to browse machines connected to your LAN in a way similar to Network Neighbourhood in the MS world. LISa allows you to browse using multiple protocols: SMB, FTP, HTTP and (allegedly) NFS. LISa has a flexible method of searching for hosts on the network. This package includes a default configuration that will browse for all hosts that offer SMB as one of their services. This will therefore pick up all Windows machines plus any boxes running Samba. To see LISa's browsing capabilities, open up Konqueror and point it at lan:/ * Sat Jun 02 2001 Michael Brown [EMAIL PROTECTED] 0.1-1mdk - First Mandrake version
[Cooker] [Contrib-RPM] squidGuard-blacklists-20010526-1mdk
Name: squidGuard-blacklistsRelocations: (not relocateable) Version : 20010526 Vendor: MandrakeSoft Release : 1mdk Build Date: Fri 01 Jun 2001 02:06:09 BST Install date: (not installed) Build Host: dolphin.home Group : System/Configuration/Networking Source RPM: (none) Size: 1368457 License: GPL Packager: Michael Brown [EMAIL PROTECTED] URL : http://www.squidguard.org/blacklist/ Summary : Blacklists for the squidGuard web filtering package Description : SquidGuard is a combined filter, redirector and access controller plugin for Squid. It is free, very flexible, extremely fast, easily installed, portable. This package includes a configuration set for squidGuard which uses the blacklists generated by the squidGuardRobot. These blacklists are entirely products of a dumb robot, and it is recommended that you review the lists before using them. * Tue May 29 2001 Michael Brown [EMAIL PROTECTED] 20010526-1mdk - First Linux-Mandrake package
[Cooker] [Contrib-RPM] cygwin-xfree86-20010525-1mdk
Name: cygwin-xfree86 Relocations: (not relocateable) Version : 20010525 Vendor: MandrakeSoft Release : 1mdk Build Date: Sat 02 Jun 2001 23:18:59 BST Install date: (not installed) Build Host: dolphin.home Group : Networking/Remote access Source RPM: (none) Size: 15444269 License: GPL Packager: Michael Brown [EMAIL PROTECTED] Summary : Cygwin/XFree86 X server for Windows Description : Cygwin/XFree86 is an X server for Windows. It enables you to open up a graphical login session to your Linux-Mandrake box. Cygwin/XFree86 is significantly slower than running a 'native' Linux X server. It is suitable for occasional or experimental use, but not for serious work. This package provides a Windows ZIP file, shared via Samba, that Windows clients can grab and install easily, without having to install Cygwin separately. Do NOT install this package if you already use Cygwin on your Windows boxes; it is likely that this package will overwrite some of your files (since Windows doesn't have a proper package management tool :-) The Samba share public will be created if it does not already exist. * Sat Jun 02 2001 Michael Brown [EMAIL PROTECTED] 20010525-1mdk - First spec file for Mandrake distribution, based on non-RPM package built by Fen Systems Ltd. on 05/25/2001 (hence version)
Re: [Cooker] [Contrib-RPM] cygwin-xfree86-20010525-1mdk
On 3 Jun 2001, andre wrote: Name: cygwin-xfree86 Relocations: (not relocateable) Summary : Cygwin/XFree86 X server for Windows Description : Cygwin/XFree86 is an X server for Windows. It enables you to open up a graphical login session to your Linux-Mandrake box. Cygwin/XFree86 is significantly slower than running a 'native' Linux X server. It is suitable for occasional or experimental use, but not for serious work. This package provides a Windows ZIP file, shared via Samba, that Windows clients can grab and install easily, without having to install Cygwin separately. Do NOT install this package if you already use Cygwin on your Windows boxes; it is likely that this package will overwrite some of your files (since Windows doesn't have a proper package management tool :-) The Samba share public will be created if it does not already exist. I have the sneaking suspicion that this will be shotdown. Besides wouldn't the java X server be better. You can than claim it's for linux *cough*. Works on any machine with a good jvm... That would mean windows and *nix :) Oh well This package is designed to help spread the use of Linux(-Mandrake). The idea is that when you have a Mandrake machine on a fast network (e.g. University) then you can let your Windows-using friends try out Mandrake without forcing them to install a complete new operating system on their boxes. Another possible situation is a Mandrake box being used as a file/proxy/etc server in a company network with Windows workstations - this package enables adventurous users to try out Linux. It's a package I've put together because I need it, and I'll be keeping it in my own private RPM repository anyway (along with other gems like my StarOffice RPM that eliminates the tedious, per-user GUI installation procedure). Someone else might also find it useful, hence I contributed it - isn't that what Contribs is for? I've already had 4/5 people asking me directly for a copy after I announced it on another mailing list, so I suspect the demand for this package may be non-zero. It doesn't bother me if it gets shotdown and doesn't end up in Contribs - just means one more package that stays in my private urpmi database. Michael
[Cooker] /incoming is full again...
Can I use the donations page to donate some money specifically to be used for buying an extra disk for /incoming? :-) Michael
[Cooker] Missing buildRequires: apache-devel and pam-devel formod_auth_external-2.1.2-5mdk
Subject says it all. Is there any way I can get CVS access so that I can put in tiny changes like this myself without having to bother the developers? Michael
[Cooker] [RPM] Uploaded to /incoming: httpd-naat-0.5-2mdk (fwd)
Name: httpd-naat Relocations: (not relocateable) Version : 0.5 Vendor: MandrakeSoft Release : 2mdk Build Date: Thu 31 May 2001 22:12:11 BST Install date: (not installed) Build Host: dolphin.home Group : System/ServersSource RPM: (none) Size: 58109License: Apache License Packager: Michael Brown [EMAIL PROTECTED] URL : http://www.linux-mandrake.org Summary : HTTP server daemon to provide Administrative WWW services Description : httpd-naat is an Apache configuration for the Network Appliance Administration Tool on Linux-Mandrake Server. * Thu May 31 2001 Michael Brown [EMAIL PROTECTED] 0.5-2mdk - Edited httpd-naat.mod_ssl.conf to reflect new location of libssl.so - Added requires mod_ssl = 2.8.2-2mdk (the earliest version in which libssl.so is in the new location). Same thing happened again: I initially tried to send this to the Packager address as well ([EMAIL PROTECTED]), but the message was rejected because: Hi Sorry, but you have tried to post a message to an internal mailing list. Only mandrakesoft employees are allowed to do so: your message will not appear on the list yours Denis Havlik Maybe [EMAIL PROTECTED] is not the best e-mail address to put on an RPM package if no-one outside of MDK can send to it...? Michael
[Cooker] Missing buildRequires: libpcap-devel for snort-1.7-3mdk (fwd)
Subject says it all. I initially tried to send this to the Packager address as well ([EMAIL PROTECTED]), but the message was rejected because: Hi Sorry, but you have tried to post a message to an internal mailing list. Only mandrakesoft employees are allowed to do so: your message will not appear on the list yours Denis Havlik Maybe [EMAIL PROTECTED] is not the best e-mail address to put on an RPM package if no-one outside of MDK can send to it...? Michael
[Cooker] [Contrib-RPM] squidGuard-1.1.4-12mdk
Name: squidGuard Relocations: (not relocateable) Version : 1.1.4 Vendor: MandrakeSoft Release : 12mdk Build Date: Thu 31 May 2001 22:41:38 BST Install date: (not installed) Build Host: dolphin.home Group : System/ServersSource RPM: (none) Size: 1342776 License: GPL Packager: Michael Brown [EMAIL PROTECTED] URL : http://www.squidguard.org Summary : Filter, redirector and access controller plugin for Squid. Description : SquidGuard is a combined filter, redirector and access controller plugin for Squid. It is free, very flexible, extremely fast, easily installed, portable. SquidGuard can be used to - limit the web access for some users to a list of accepted/well known web servers and/or URLs only. - block access to some listed or blacklisted web servers and/or URLs for some users. - block access to URLs matching a list of regular expressions or words for some users. - enforce the use of domainnames/prohibit the use of IP address in URLs. - redirect blocked URLs to an intelligent CGI based info page. - redirect unregistered user to a registration form. - redirect popular downloads like Netscape, MSIE etc. to local copies. - redirect banners to an empty GIF. - have different access rules based on time of day, day of the week, date etc. - have different rules for different user groups. Neither squidGuard nor Squid can be used to - filter/censor/edit text inside documents - filter/censor/edit embeded scripting languages like JavaScript or VBscript inside HTML * Tue May 29 2001 Michael Brown [EMAIL PROTECTED] 1.1.4-12mdk - Split package into main and sampleconfig to allow other squidGuard configuration packages (provides: squidGuardConfig) to exist - Added code to symlink /etc/squid/squidGuard.conf to sample config file in post and preun of sampleconfig package - Added automatic handling of /etc/squid/squid.conf in post and preun - Added call to service squid reload in post and preun - sampleconfig Requires: httpd-naat - Modified squidGuard.conf.sample to reflect the cgi that is in the httpd-naat package (squidGuard.cgi instead of squidGuard-simple.cgi) - Removed rm -rf from postun - causes errors on removing packages because squidGuard deletes squidGuard-sampleconfig's files
Re: [Cooker] Segfault/lockup: Gimp / libeazel-engine.so
On Tue, 29 May 2001, Frederic Crozat wrote: There seems to be something wrong with libeazel-engine.so. On machine 1 (AMD Athlon, 256MB RAM) it will segfault most of the time if you close a window and then select a new tool. Output of gdb is: It's a know problem.. I'm working on it.. Excellent. Will it appear as an update for 8.0 or only in cooker? Michael
[Cooker] Exclusive RPMS?
Is there any way to set up spec files so that only one RPM in a group can be installed. I'm thinking along the lines of: Each of the packages foo-abc, foo-def and foo-xyz provides foo Only one of packages foo-abc, foo-def and foo-xyz should be able to be installed at any one time. From memory, if I put Provides: foo, Conflicts: foo, then all packages will refuse to install. I remember this being discussed in relation to the Aurora Monitors, but I've looked at the Aurora spec file and it doesn't seem to implement anything that looks relevant. Any ideas? Michael
[Cooker] [Contrib-RPM] Uploaded to /incoming: lilypond-1.4.2-1mdk
Name: lilypond Relocations: (not relocateable) Version : 1.4.2 Vendor: MandrakeSoft Release : 1mdk Build Date: Tue 29 May 2001 21:07:28 BST Install date: (not installed) Build Host: dolphin.home Group : PublishingSource RPM: (none) Size: 1077567 License: GPL Packager: Michael Brown [EMAIL PROTECTED] URL : http://www.lilypond.org/ Summary : A program for printing sheet music. Description : LilyPond is a music typesetter. It produces beautiful sheet music using a high level description file as input. Lilypond is part of the GNU project. LilyPond is split into two packages. The package lilypond provides the core package, containing the utilities for converting the music source (.ly) files into printable output. The package lilypond-extras provides the full documentation, example .ly files for various features and the Mutopia project files (musical equivalent of the Gutenberg project - see http://www.mutopiaproject.org for details). If you are new to lilypond, you will almost certainly want to install the lilypond-extras package in addition to the lilypond package. You may also wish to investigate the denemo package, which provides a graphical front end to lilypond. See the file README.first for more information. * Mon May 28 2001 Michael Brown [EMAIL PROTECTED] 1.4.2-1mdk - Upgraded to 1.4.2 - Changed URL - Removed /etc/profile.d scripts: all TeX bits are now located properly - Rearranged documentation - Tidied spec file - Updated README.first
[Cooker] Segfault/lockup: Gimp / libeazel-engine.so
There seems to be something wrong with libeazel-engine.so. On machine 1 (AMD Athlon, 256MB RAM) it will segfault most of the time if you close a window and then select a new tool. Output of gdb is: GNU gdb 5.0mdk-11mdk Linux-Mandrake 8.0 Copyright 2001 Free Software Foundation, Inc. GDB is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. Type show copying to see the conditions. There is absolutely no warranty for GDB. Type show warranty for details. This GDB was configured as i386-mandrake-linux...(no debugging symbols found)... (gdb) run Starting program: /usr/bin/gimp (no debugging symbols found)...(no debugging symbols found)...(no debugging symbols found)...(no debugging symbols found)... (no debugging symbols found)...(no debugging symbols found)...(no debugging symbols found)...(no debugging symbols found)... (no debugging symbols found)...[New Thread 1024 (LWP 19239)] Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 1024 (LWP 19239)] 0x4046c6c7 in eazel_engine_fill_gradient_rgb_buffer () from /usr/lib/gtk/themes/engines/libeazel-engine.so (gdb) bt #0 0x4046c6c7 in eazel_engine_fill_gradient_rgb_buffer () from /usr/lib/gtk/themes/engines/libeazel-engine.so Cannot access memory at address 0xc002df80 (gdb) On machine 2 (Twin PIII, 512MB) it locks up (100% CPU time) when changing tools. Gdb shows that it is stuck inside the same routine (eazel_engine_fill_gradient_rgb_buffer). Versions: gimp-1.2.1-5mdk mandrake_desk-8.0-8mdk Michael
[Cooker] [Contrib-RPM] ucblogo-5.1-1mdk
Name: ucblogo Relocations: /usr Version : 5.1 Vendor: MandrakeSoft Release : 1mdk Build Date: Mon 28 May 2001 19:31:48 BST Install date: (not installed) Build Host: dolphin.home Group : Development/Languages Source RPM: ucblogo-5.1-1mdk.src.rpm Size: 1277089 License: GPL Packager: Michael Brown [EMAIL PROTECTED] Summary : Berkeley Logo interpreter Description : Berkeley Logo interpreter for Unix and X. Features *not* found in Berkeley Logo include robotics, music, GUIs, animation, parallelism, and multimedia. For those, buy a commercial version. * Mon May 28 2001 Michael Brown [EMAIL PROTECTED] 5.1-1mdk - Upgraded to 5.1 - Tidied spec file - Removed now unnecessary parse.c and main.c patches - Added BuildRequires for libtermcap-devel - Added call to makeinfo to regenerate old info files - Patched texi source to include info-dir listing - Added install-info - Removed libloc.c to trigger correct path usage - Added hack to move Messages into logolib folder (seems to be missing from makefile) - Added icon and menu entry * Sat Feb 10 2001 Michael Brown [EMAIL PROTECTED] 4.6-1mdk - First spec file for Mandrake distribution - Patched parse.c and main.c to correct initialiser not constant error for {read,write,load}stream
[Cooker] [Contrib-RPM] rdesktop-1.0.0.pl19.6.1-2mdk
Name: rdesktop Relocations: /usr Version : 1.0.0.pl19.6.1Vendor: MandrakeSoft Release : 2mdk Build Date: Mon 28 May 2001 19:52:00 BST Install date: (not installed) Build Host: dolphin.home Group : Networking/Remote access Source RPM: rdesktop-1.0.0.pl19.6.1-2mdk.src.rpm Size: 114725 License: GPL Packager: Michael Brown [EMAIL PROTECTED] URL : http://www.rdesktop.org/ Summary : rdesktop is a terminal server client for Windows NT4 and Windows 2000 Description : rdesktop is an open source client for Windows NT Terminal Server and Windows 2000 Terminal Services, capable of natively speaking their Remote Desktop Protocol (RDP) in order to present the user's NT/2000 desktop. rdesktop currently runs on Linux and most other UNIX based platforms with the X Window System. * Thu May 10 2001 Michael Brown [EMAIL PROTECTED] 1.0.0.pl19.6.1-2mdk - Added BuildRequires: openssl-devel * Wed May 09 2001 Michael Brown [EMAIL PROTECTED] 1.0.0.pl19.6.1-1mdk - First Mandrake release
[Cooker] [Contrib-RPM] Missing BuildRequires: octave-2.0.16-8mdk
Needs a BuildRequires: egcs-c++ since egcs is not the standard c++ compiler. Michael
[Cooker] DrakX v1.510 bug report: printer name
Configuring printer during installation (DrakX v1.510, Mandrake 8.0 plus current updates, ran gendistrib). Printer name does not accept the | character which is used to specify multiple names for the same printer (according to the help text displayed with the dialog during installation). Suggested patch (not tested): --- printerdrake.pm Wed Apr 18 09:19:14 2001 +++ printerdrake.pm.new Mon May 28 01:35:10 2001 @@ -450,8 +450,8 @@ { title = _(Select Printer Connection), cancel = !$::expert || !$printer-{configured}{$printer-{QUEUE}} ? '' : _(Remove queue), callbacks = { complete = sub { -unless ($printer-{QUEUE} =~ /^\w*$/) { -$in-ask_warn('', _(Name of printer should contains only letters, numbers and the underscore)); +unless ($printer-{QUEUE} =~ +/^\w+[\w\|]*\w+$/) { +$in-ask_warn('', _(Names of printer +should contains only letters, numbers and the underscore and should be separated by +|)); return (1,0); } return 0; Michael
Re: [Cooker] XFree locks up computer when XFdrake
On Fri, 25 May 2001, SI Reasoning wrote: Unfortunately that problem involved the mouse being locked up as well as the keyboard. However that does sound interesting. How would I get gpm to set up these keystrokes and associations? Maybe it should be a standard part of the mandrake distribution to have such an out! It was probably X that locked up, rather than the mouse hardware. If it had been routed via gpm to X, then gpm should still have been able to pick up the emergency sequence even if X had stopped responding to mouse movements. To get gpm to act as a repeater, change /etc/sysconfig/mouse to contain MOUSETYPE=xxx -R msc where xxx is your real mousetype (see gpm manpage for a listing), then tell X that it has ProtocolMouseSystems Device /dev/gpmdata The gpm manpage tells you how to set up the emergency sequences. HTH, Michael
Re: [Cooker] XFree locks up computer when XFdrake
On Fri, 25 May 2001, SI Reasoning wrote: I cannot get XFdrake to properly read my system.with the new XFree86. First it does not give the 3.3.6 or 4 choice anymore. Then after it asks questions and I set it to config that I always do, it will totally lock up the system when doing the test until I have to hard reboot. Immediately upon X the first thing I notice is that the mouse does not work, then I notice that nothing else does either including our favorite ctrl + Alt + BS.Often after such a hard lock, the system will kick out to maintenance where I have to do a manual fsck. Not a solution, but something that might come in handy: Executing the command /usr/bin/kbd_mode -a will take away XF86's control of the keyboard, and you can then use Alt-F1 to get back to a console and find out what's going wrong with X. If you use gpm's repeater mode (rather than letting X talk to the mouse directly), then you can use gpm's Special commands feature which enables you to associate commands with unlikely mouse sequences (such as triple-clicks). You can therefore set up gpm to execute /usr/bin/kbd_mode -a if you e.g. triple-click both buttons then wait for a beep and click the left button again, thus giving you a handy escape route if XF86 freezes up. Incidentally, I've found that mouse behaviour in X is a lot better when routed via gpm - X refused to acknowledge the middle button on some mice but gpm picked it up happily and forwarded it to X in a way that X understood. Using gpm also means that you can plug the mouse in after starting X, or even change from a serial to a PS/2 mouse in the middle of an X session. HTH, Michael
Re: [Cooker] Eazel services picked up and expanded by Mandrake?
On Tue, 22 May 2001, Armisis Aieoln wrote: I think what he was talking about was... instead of going public with an ipo, have mandrake be a user/developer owned company with the ability to buy into the company that you are also putting effort into. Sounds like a good idea to me. What is the legal status of Mandrake as a company? If it's the equivalent of what in the UK is a private limited company, then it could offer shares for sale to developers (but not to the general public), AFAIK. Now this is sounding intresting Is anyone at Mandrake seriously considering this? I don't know any French law, but if the legal status is similar to that of a UK private limited company then Mandrake can offer shares for sale to whoever it wants to on an individual basis, but cannot open them up for sale to the general public. I can't speak for anyone else, but I would be keen to buy a few shares every so often (say $10 a week - $500 a year) and know that I was helping to keep Mandrake alive and at the same time building up an investment (donations aren't quite the same) ready for the time when Mandrake displaces Windows on the majority of desktop PCs(!). Thoughts? Michael
Re: [Cooker] Eazel services picked up and expanded by Mandrake?
On Sun, 20 May 2001, SI Reasoning wrote: I think what he was talking about was... instead of going public with an ipo, have mandrake be a user/developer owned company with the ability to buy into the company that you are also putting effort into. Sounds like a good idea to me. What is the legal status of Mandrake as a company? If it's the equivalent of what in the UK is a private limited company, then it could offer shares for sale to developers (but not to the general public), AFAIK. Michael
Re: [Cooker] Eazel services picked up and expanded by Mandrake?
On 17 May 2001, Guillaume Cottenceau wrote: All IT and Linux companies seem to be facing the same situation, currently. Other than the obvious (buying boxed sets, donating directly), what are the ways that we can help generate extra income? Thanks :-). Actually I don't know. I think the best would be to buy the boxed 8.0 from our website. It's priced $69 + $20 for postage and support of development. I think a larger part of this money comes to us than buying in compUSA. Other than that, I know donation. But the rest... I'm not sure :-). Praying? :-) Seriously, I think buying the boxed version (or give 1/3 of $$ as donation) would be a great idea to support our efforts. Well, I've done the donation thing. One idea I've had: is Mandrake going to be present at the Linux Expo in Birmingham, UK in September? If not, then I'll be there anyway, as part of a non-profit, Linux-in-education group with two demonstration machines running Mandrake, and I could possibly sell boxed sets to anyone who asked where they could get a copy of this distribution that's so much better than anything else in the Expo! Michael
Re: [Cooker] Eazel services picked up and expanded by Mandrake?
On 17 May 2001, Guillaume Cottenceau wrote: I liked the drag and drop interface for moving files within eazel services. What if Mandrake picked it up and allowed for such things as photo galleries, etc like Yahoo. Maybe Mandrake can have a linux enhanced web portal... with email , calendering, etc... that sync's with netscape, outlook, and the gnome and kde apps like calendering, etc. There is a strong need for these things and it could be another potential revenue for Mandrake. Currently we're looking for actual revenue rather than potential revenue since we're facing very hard times in terms of cash. Does this mean that my favorite distro may not be around much longer? I don't know. All IT and Linux companies seem to be facing the same situation, currently. Other than the obvious (buying boxed sets, donating directly), what are the ways that we can help generate extra income? Michael
Re: [Cooker] Common Linux Installer
On 14 May 2001, Guillaume Cottenceau wrote: snip Also, we strongly support the fact that our installer is written in Perl language, while that choice would not fit Redhat or Debian interestes for sure. Just wanted to say that I am a great fan of the Perl installer - it makes it so much easier to debug auto_installs. Michael
Re: [Cooker] gendistrib requirements?
On 14 May 2001, François Pons wrote: snip Thanks again, and I hope that the patch makes sense (I'm not really a Perl programmer). Your patch make sense, this is because of that a shell is used. I will modify rpm2header to make it more usable for older distro so that using misc/gendistrib (for example) will make sure no patch are necessary as so shell will need to be invoked (no more need to use , so no more shell). Thanks. BTW, I think that the use of the pipe through packdrake also invoked a shell, because it broke on that line as well. Michael
Re: [Cooker] OpenOffice: Sun JDK spec file?
On Mon, 14 May 2001, Gwenole Beauchesne wrote: I would like to bring OpenOffice up to date (the latest release does include some vital features such as printing) unless the current maintainer (fcrozat) objects. OpenOffice 627 needs fixes in order to comply with the ISO C++ standard. Currently, it does not compile with gcc-2.96 nor with gcc-3.0 (20010511). Work is in progress. Is this work is in progress within MandrakeSoft or work is in progress within Sun? I ask merely because I am keen to evaluate OpenOffice, and if it doesn't have a high priority within MandrakeSoft then I'm quite prepared to put in the effort myself. At the same time, I don't want to waste my time if it turns out that there are already people working on it. Michael
[Cooker] OpenOffice: Sun JDK spec file?
I would like to bring OpenOffice up to date (the latest release does include some vital features such as printing) unless the current maintainer (fcrozat) objects. One of the build requirements is Sun's JDK. I would like to make sure that my installation of JDK matches the RPM'd version, so that I can be sure my resulting SRPM will build at MandrakeSoft. Can someone point me towards a copy of the .spec file used by Mandrake? (I have already looked in CVS in the SPECS/ folder and cannot find it). TIA, Michael Brown
[Cooker] gendistrib requirements?
Just tried to run gendistrib to build the 8.0 hdlists etc. 8.0 tree is stored on a machine running 7.2. On running gendistrib, I get: Can't load './auto/rpmtools/rpmtools.so' for module rpmtools: librpmio.so.0: cannot open shared object file: No such file or directory at /usr/lib/perl5/5.6.0/i386-linux/DynaLoader.pm line 200. at ./gendistrib line 20 According to rpmfind.net, librpmio.so.0 is provided by rpm4, which explains why the 7.2 system (with rpm 3.0.5) doesn't have it. This is a chicken-and-egg situation: I need rpm4 (part of 8.0) to build the 8.0 distribution media. Is there any way that librpmio could be provided along with rpmtools in /misc? TIA, Michael Brown
Re: [Cooker] gendistrib requirements?
On Fri, 11 May 2001, Blue Lizard wrote: Blue With gendistrib, is it possible to contribs to rpms3 (as rpms2 is on second cd/iso) and have 3rd cd that works just like 1 2 with the auto dep checking and descriptions, etc.? I assume it was made for stuff like that. Or if installdone from hd, take all rpms to one dir and gendistrib and make a massive (far too big for cd) iso for use off hd? /Blue Yes. For the first: just create an entry in the Mandrake/base/hdlists file that points to your other RPM repositories, then re-run gendistrib. For the second: move the RPMS from RPMS2 into RPMS, edit Mandrake/base/hdlists appropriately (i.e. remove reference to RPMS2) and re-run gendistrib. I'm using it to generate installation media that contain a few contrib RPMS and all the updates to 8.0. Michael
[Cooker] Uploaded to /incoming: rdesktop-1.0.0.pl19.6.1-1mdk.src.rpm
Name: rdesktop Relocations: (not relocateable) Version : 1.0.0.pl19.6.1Vendor: MandrakeSoft Release : 1mdk Build Date: Thu 10 May 2001 02:10:07 BST Install date: (not installed) Build Host: dolphin.home Group : Networking/Remote access Source RPM: (none) Size: 66285License: GPL Packager: Michael Brown [EMAIL PROTECTED] URL : http://www.rdesktop.org/ Summary : rdesktop is a terminal server client for Windows NT4 and Windows 2000 Description : rdesktop is an open source client for Windows NT Terminal Server and Windows 2000 Terminal Services, capable of natively speaking their Remote Desktop Protocol (RDP) in order to present the user's NT/2000 desktop. rdesktop currently runs on Linux and most other UNIX based platforms with the X Window System. * Wed May 09 2001 Michael Brown [EMAIL PROTECTED] 1.0.0.pl19.6.1-1mdk - First Mandrake release
Re: [Cooker] Java in Mdk8
On Sat, 5 May 2001, Alexander Skwar wrote: Nothing depends on StarOffice rpm, but all java rpm depends on jre or jdk rpm.. That makes a difference for me. Exactly. And installing StarOffice is very easy and the way it installs does away with the need for a RPM. Unless you want to install it on many different machines, in which case RPM means a lot less work... Michael