Re: amd64 situation in Mirrors.masterlist
Goswin von Brederlow wrote: Should the master list ignore the sarge has no amd64 issue and just list amd64 for mirrors that have it in etch/sid? Or should there be two entries for every mirror, one for sarge/!amd64 and one for etch/sid with all archs? Mirrors.masterlist does not provide a way to limit a mirror to a given release. It's only used by d-i for etch anyway, and for the web site. On the same note, should the sarge/amd64 mirrors get removed in preparation of stable/etch having amd64? Those mirrors have already been removed. -- see shy jo signature.asc Description: Digital signature
Re: amd64 situation in Mirrors.masterlist
Jo Shields wrote: Would it be useful to refine the rather rushed tools used to create the diff, removing any manual steps from Mirrors.masterlist to .diff? Yes. -- see shy jo signature.asc Description: Digital signature
Re: amd64 situation in Mirrors.masterlist
Jo Shields wrote: Okay then. Attached is the new, improved Masterlister program. It should just compile with mcs Masterlister.cs, using the MCS compiler from Mono. Usage is Masterlister.exe path/to/source/masterlist - a diff is produced on STDOUT (with program info on STDERR, a bit like OggEnc). I guess the really nice thing to do would be to integrate this with http://www.de.debian.org/dmc/today/ Or it could be checked into the website cvs, though it would probably be best to put it under an actual defined license like the GPL first. Also attached is another diff. Applied. -- see shy jo signature.asc Description: Digital signature
Re: amd64 situation in Mirrors.masterlist
Andreas Barth wrote: * Jo Shields ([EMAIL PROTECTED]) [060808 12:17]: What's the correct procedure for a) Defunct, unresponsive or incorrect mirror entries such as 'Site: www.zentek-international.com' (wrong Archive-http) or 'Site: natasha.stmarytx.edu' (timeout) b) the six mirrors with sh architecture? Prodding the administrators, and, failing that, removing the entry. And bearing in mind that one or two mirrors are only accessible from inside their target country. (Which certianly makes it hard to check them..) -- see shy jo signature.asc Description: Digital signature
Re: amd64 situation in Mirrors.masterlist
Jo Shields wrote: I'm not actually sure what the correct format is for Archive-architecture when something's got all supported arches The best thing to do there is to list all the arches. and I'm not sure what to do about the hosts I couldn't contact - but this should save someone some time. **BADMIRROR** lines need manual checking, I'm afraid. A patch that updates Mirrors.masterlist with the arch info for the mirrors you could reach would be helpful. -- see shy jo signature.asc Description: Digital signature
amd64 situation in Mirrors.masterlist
A quick look at what Mirrors.masterlist says about mirrors carrying amd64 finds a lot of probably wrong information: * 154 mirrors listed as mirroring !amd64, but having all other arches. This is probably wrong for most of them ... it's definitly wrong for the couple I spot-checked. * 69 mirrors listed as mirroring only amd64 and one or two other arches. Probably correct for most of them. * 65 mirrors listed as mirroring some list of archirectures that does not include amd64. Such as mirroring only i386. Probably right for many and wrong for many others. Would someone like to go through and fix all this in Mirrors.masterlist? Bonus points if you fix the archive lists to be accurate for all other arches too. Super bonus points if you don't do it by hand like I did, last time I did it. :-) (In case you're not familier with Mirrors.masterlist, it is here: http://cvs.debian.org/webwml/english/mirror/Mirrors.masterlist?rev=1.636root=webwmlview=log And it controls what mirrors users are shown for their arch in d-i, amoung other things.) -- see shy jo signature.asc Description: Digital signature
Re: Debian Installer - Etch Beta 3 release
Frederik Schueler wrote: I would like to upload kernel-wedge 2.24 before creating new amd64 udebs, as there have been some changes to k-w which are needed for the new udebs. The k-w changes are specific to the nic-extra-modules udeb, and might touch other architectures too. can I just upload it tomorrow in order to make dinstall or are other changes pending? It only affects amd64, alpha, and i386. I believe all three should work with the new kernel-wedge. I've already uploaded the new version. -- see shy jo signature.asc Description: Digital signature
Re: Dropping the amd64-generic flavour
Otavio Salvador wrote: Frederik Schueler [EMAIL PROTECTED] writes: Please point me to other places, where this will cause breakage, and I will help fixing it. AFAIK, you'll need to fix base-installer too. Also debian-cd. -- see shy jo signature.asc Description: Digital signature
Re: Bug#367634: Fwd: support for Symbios Logic 53c1030 PCI-X
Frederik Schueler wrote: the daily d-i images do not build for amd64, because of testing being in a pretty bad shape, many d-i copmponents are still missing. The last available daily image was built using the old, unofficial archive. This no longer seems to be the case; I don't see any missing udebs in testing for amd64 now. Also, xorg is in testing on amd64. Can we restart the builds? -- see shy jo signature.asc Description: Digital signature
Re: The future of the amd64 port
Goswin von Brederlow wrote: Can we list the amd64.debian.net mirrors as sarge only and the debian mirrors as etch/sid? Not sure if the Master file had that info. The sarge installer contains a copy (or 2) of the mirror list, so changing Mirrors.masterlist for etch will not affect sarge, aside from needing to be careful that any new versions of the sarge packages avoid downloading the updated list. (Remember to cc me.) -- see shy jo signature.asc Description: Digital signature
Re: beta status
Kurt Roeckx wrote: Rejected: debian-installer-images_20051026_amd64.tar.gz: changes file doesn't list `source' in Architecture field. Second try.. this seems like it's expecting a sourceful upload for some reason, which it shouldn't if you have the debian-installer source package already in the repository from the main debian repo. -- see shy jo signature.asc Description: Digital signature
Re: beta status
Kurt Roeckx wrote: The next problem is to actually get it into the archive. It's getting rejected: Rejected: debian-installer-images_20051026_amd64.tar.gz: changes file doesn't say debian-installer-images_20051026 for Source Rejected: debian-installer-images_20051026_amd64.tar.gz: should be 20051026 according to changes file. Rejected: debian-installer-images_20051026_amd64.tar.gz: changes file doesn't list `source' in Architecture field. It also did that with the previous version that build (20051009), and I have no idea why. It seems to think that debian-installer-images_20051026_amd64.tar.gz is the source while it's actually a (special) binary package. Sounds to me like a broken .changes file.. -- see shy jo signature.asc Description: Digital signature
Re: beta status
Kurt Roeckx wrote: I would be suprised if it generated broken .changes files. I've attached it. It's broken, what's the listed debian-installer_20051026_amd64.deb? Files: cf4dca5ef36c3ab9e8794be42533fd5e 574138 devel optional debian-installer_20051026_amd64.deb 35a90f59e9d7ee2694ad2e94e1b30746 49826709 raw-installer - debian-installer-images_20051026_amd64.tar.gz Should list only the second of these. -- see shy jo signature.asc Description: Digital signature
Re: beta status
Kurt Roeckx wrote: We've pushed apt 0.6.42.2 to testing in the amd64 archive. We didn't have the problem with the gcc-4.0 dependency and did a local override to get the new apt in testing. That's very good (and timely) news. Is there any status on getting the debian-installer package built? -- see shy jo signature.asc Description: Digital signature
Re: beta status
Kurt Roeckx wrote: The problem now seems to be that rootskel-locale still seems to exist in testing for some reason. It's unclear to me why it still exists. This is causing the monolithic target to fail to build because it can't find the locale. I've asked the amd64 ftp-master too look at it, but it's currently still not solved. They've added it to our reject file (like that did for partman), but in this case it doesn't seem to have helped. I hope they find the problem soon so I can get this build. You know, amd64 is the only arch to build the monolithic target by default. I think that this is because it used to be hard to get businesscard CDs for amd64, but we build them now. And also there used to be the mirror selection issues which made it harder to install amd64 than other arches, but those are also now resolved. So you could just turn off this target. It's mostly only useful for debugging non-uploaded udebs during development. -- see shy jo signature.asc Description: Digital signature
Re: beta status
An update on the d-i beta status. We're getting really close, AKA most things seem likely to work now. Ccing some other relevant lists. debian-boot: - Thanks to fjp, base-installer 1.35.4 should get d-i working again with secure apt and CDs, but we're currently mssing uploads of successful builds for 3 architectures. This is the last udeb we plan to put into testing for the beta, once it's built everywhere. Also, once this udeb does reach testing, it should be possible to do some etch_d-i CD installs and test things out. The fixed base-installer will reach testing with today's mirror sync. So within an hour or two (netboot etc) and after tonight's build (CDs) the etch d-i images can be used to test the beta and should actually work. Your testing and reports are appreciated, as we decide when to make the beta final. Some links for those images: floppy, netboot, etc: http://ftp.us.debian.org/debian/dists/sid/main/installer-$ARCH/ CD: http://cdimage.debian.org/pub/cdimage-testing/etch_d-i/ This is just a beta so I am not going to be too picky about testing, but it would be nice to fill out as much of installer/doc/devel/release-checklist as we can. At a minimum we need to make sure that businesscard, netinst, and full CDs (once we get some) work for i386 and powerpc and that the desktop task installs ok and works. alpha: - debian-installer FTBFS on alpha, but apparently only on the buildd. http://buildd.debian.org/fetch.php?pkg=debian-installerver=20051026arch=alphastamp=1130532682file=logas=raw We need this build if alpha will be in the beta. The other builds of 20051026 should be final for the beta, but still need to be installed and tested out. This is still a problem and I've seen no progress on this issue. Even someone doing a manual build and upload on alpha would probably be acceptable this point, (as long as you file a FTBFS bug too or something so we remember to investigate the buildd issue later..). amd64/debian-release: - amd64 CDs seem to be significantly broken, we've been getting many failure reports all week. (#336353, #335556, #335653, #336173, #336451) Unless this is resolved and we see some successful amd64 installs, it won't be in the beta. This was resolved, only to hit the next problem with amd64: The amd64 archive signing key is not trusted by apt. So currently testing amd64 installs only work from the netinst CD, all the other install methods, which use apt authentication, are broken. This is fixed in apt 0.6.42.2, but it won't reach testing in a while due to annoying gcc-4.0 dependencies needing to reach testing first. amd64 has also not built the most recent version of the debian-installer package, and has been marked as building for over a day at http://people.debian.org/~igloo/status.php?email=packages=debian-installerarches= Additionally, it seems that the last debian-installer build to be built and installed into the amd64 archive was rc3, in May. See http://amd64.debian.net/debian/dists/sid/main/installer-amd64/ So I'm not even sure if version 20051026 will get properly installed even if it does get built for amd64. At this point I'm not sure what to do about amd64 and the beta. I would rather not wait for a possibly indefinite gcc-4.0 transition to get the new apt in. Only supporting the amd64 netinst could work, so could doing some magic to get an upated apt into testing. m68k: I've become aware of another issue, which is that some m68k d-i udebs were miscompiled by a broken compiler there and don't work. I understand that smarenka has been working on this, but I don't know the currently status of it (beyond what's documented at http://wiki.debian.org/DebianInstallerM68kTodo) and whether m68k will be included in the beta is uncertian. -- see shy jo signature.asc Description: Digital signature
Re: SATA install - AMD64 - success story
Lennart Sorensen wrote: I think if Lenart Sorensen remasters an AMD64 installer with a 2.6.12 kernal *and* compiles the AHCI module (it's under drivers/scsi) then I'll know what to do.[1] Oh now you tell me. Off to rebuild again. It is enabled in the kernel, bit I didn't add it to the list of modules to include... I just did now. ahci has been included in sata-modules by kernel-wedge if it's available since May of this year, and is in the standard installer images using 2.6.11 and the not-yet-uploaded ones using 2.6.12. -- see shy jo signature.asc Description: Digital signature
Re: amd64 netinst
Santi wrote: I have installed succesfully debian amd64 from the netinst CD. Thanks for the note. I've set its state to working on the web site. I still wonder how the amd64 netinst manages to be 69 mb when 32 bit i386 has a 110 mb one. Something *must* be missing.. -- see shy jo signature.asc Description: Digital signature
Architecture lines added to Mirrors.masterlist
I've extended the format of the the Mirrors.masterlist file, which is used to record available Debian mirrors, to include machine-readable information about what architectures are included in a mirror. This information will be used by the Debian installer and other programs to select an appropriate mirror for a user's architecture. I'm ccing this to the debian-mirrors list because I suspect that a lot of our information about which mirrors are mirroring which architectures is out of date. The information about which architectures is available is now included in the list of mirrors at http://www.nl.debian.org/mirrors/list, so please check that the information for your mirror is accurate, and if not, please let us know. If you change the architectures included in your mirror in the future, we'd appreciate a mail mentioning this so we can keep the information up-to-date. Also, note that I have added the amd64 Mirrors.list file onto Mirrors.masterlist, but at least for now the amd64 information is filtered out of the mirror list page on the official Debian website. If it's decided that should be included there already, it's very easy to remove the filter. The amd64 people may want to switch over to using Mirrors.masterlist directly, or should at least keep us current as they continue adding mirrors. -- see shy jo signature.asc Description: Digital signature
Re: Bug#282763: Installation report amd64 (gcc-3.4)
Harald Dunkel wrote: Joey Hess wrote: I thought EFI was only an ia64 thing. Strange. Could you send a tarball of /var/log/debian-installer/ from the installed system so I can try to see why it was doing EFI stuff? See attachment. From the d-i status file: Package: partman-efi Status: install ok unpacked Version: 7.0.0.1.amd64 Depends: parted-udeb, partman Description: Add to partman support for EFI boot partitions I conclude that the unofficial amd64 archive is broken or someone is doing something strange putting this udeb in it. And what's up with the .amd64 version numbering? Note that the current released version of this package is 6. -- see shy jo signature.asc Description: Digital signature
Re: Bug#279508: Debian AMD64 installer error
gulfstream wrote: When the Debian AMD64 installer load SCSI driver module, the installer crashed. My motherboard is Tyan S4880, CPU are 4 Opteron 848. The SCSI card is SLI 53C1030 PCI-X Fusion-MPT Dual Ultra320 SCSI, which embed on motherboard. Please describe the crash in as much detail as possible. -- see shy jo signature.asc Description: Digital signature
Re: installing on ASUS K8N-E Deluxe
Christian T. Steigies wrote: The i386 install had one major obstacle, the onboard network was not detected, it might have loaded the nic module, but DHCP and manual network config did not work. I think this is due to a second network card magically appearing on the box, ethernet over firewire, which received eth0, the real network card received eth1. After changing network/interfaces to use eth1 instead of eth0, everything worked out smootly. The firewire ethernet will always be available as eth0 on systems running 2.6 that have a firewire controller. d-i should not have defaulted to a probably non-fuctional firewire interface if there was another ethernet interface that had link. If it did default to eth0 in the question about which one to use, this might be a bug in the link detection code. The AMD64 install had another major obstacle, the installer did not find the release file. It was looking in (typing from memory) debian/dists//Release, Most times I've seen this reported it's been due to a problem reading the CD, which makes it not notice a symlink and not figure out what the suite is. We've mostly seen this one much older machines though, with DMA problems. -- see shy jo signature.asc Description: Digital signature
Re: installing on ASUS K8N-E Deluxe
Christian T. Steigies wrote: There was nothing hooked up to the firewire, but the NIC was connected to a switch. So I guess it is a bug in the link detection code. Could there be anyhting in the installer log files? I could try to retrieve them tomorrow. I also tried a knoppix 3.6 (i386), it did not detect the network either. Yes there should be some info in there about the link info it reported. We probably need that and the model of nic card to do more. -- see shy jo signature.asc Description: Digital signature