While attempting an install without network using CDs burned from the 1.4 final pentium3 ISO images, the installation process fails to validate genkernel and decides to try to download the package. It also does this for gentoo-sources and all of the system logger tbz2 files. Testing the genkernel-1.5.tbz2 file with `tar -tjvf genkernel-1.5.tbz2` gives the following error for usr/sbin/genkernel "bzip2: (stdin): trailing garbage after EOF ignored". The file /usr/sbin/genkernel cannot be extracted. The emerge utility does not try to extract it because the md5sum that the portage system expects from the file does not match for the .tbz2 file. Reproducible: Always Steps to Reproduce: 1. Install using pentium3 ISO images. (d6004e2f8b383c0aed4f9c5cfa1381e4 pentium3-1.4-20030801-cd1.iso) 2. At the point of installing the kernel and genkernel, the system will attempt to load the source tar.bz2 files from the network. 3. Compare the md5sum of genkernel-1.5.tbz2 to the portage tree's expected value. Actual Results: I was unable to install the kernel, genkernel, and system logger without network access to download intact .tar.bz2 files. Expected Results: I should have been able to install a kernel, genkernel, and system logger using the source tarballs on the GRP CDROMs. I burned the CDs twice thinking that I might have had a problem. The ISO images produce the correct md5 checksums, and cdrecord reports no problems. The cdrecord buffer minimum buffer fill percentage was 81%, so I don't think any CD burning problems are involved.
Extracting the genkernel tbz2 file from the md5sum-validated ISO CD image shows that it is corrupt, so it is definately not a CD-burning issue.
FYI, I also observed similar behaviour on my i686 LiveCD when I was trying to install. With no network connection, "emerge -k genkernel" attempted to download sources. However, this was because the .tbz2 image in "packages" was genkernel-1.5, whereas the latest .ebuild in /usr/portage/sys-kernels/genkernel was genkernel-1.2. "Cp"ing genkernel-1.2 to genkernel-1.5 caused the "emerge -k" to succeed. It would appear that the portage tree and packages directory are not in sync on the LiveCD Images.
The 1.4 maintenance release 1 will have portage snapshots that are 100% in-sync with the tbz2 collection. But this points out a strange portage bug. Even with genkernel-1.2 in /usr/portage and genkernel-1.5 available as .tbz2, emerge -k *should* have used genkernel-1.5. So I'm forwarding this bug over to Nick.
This also happens with the pentium4 ISO's. I also had this problem with the pentium3 ISO's. I have two machines I was attempting this on with each type of chip. I did not have a problem with gentoo-sources, or genkernel. However, I did have this problem with hotplug, metalog (actually any of the system loggers don't work), and vcron (actually any of the CRON Daemons don't work). WORKAROUND (Potentially dangerous, not really sure): # emerge -k /usr/portage/packages/ALL/desired_package.tbz2 This worked and solved the dependency problem. However, I found it odd that hotplug depended on an *rc3 usbutils, packages had an rc1, once I did the workaround, hotplug no longer had the dependency even though I installed an older usbutils? This is why I mention the hotplug as dangerous. John Pywtorak Cal Poly jpywtora@calpoly.edu
well, too late now, 2004.0 is about out.
