Andres Salomon wrote: > > Currently there is a very nasty trap for anyone upgrading a system that > > relies on ndiswrapper for its networking from stable to testing. > > What nasty trap is that? The fact that the version in testing is so > old?
I upgraded everything, which included switching to udev, removing hotplug etc and pretty well broke running the system with the 2.6.8 kernel. Then I booted to the new 2.6.15 kernel, and ran module-assistant to build ndiswrapper modules for it, and found that it FTBFS, at which point I was left without a wireless network to upgrade the ndiswrapper versions to unstable. I was able to boot back to 2.6.8 and luckily the kernel was still working to the point of getting me on the network (alhough X wasn't working, and in some cases I imagine the system wouldn't be usable at all without some fairly involved handholding). So then I upgraded the ndiswrapper-source, booted back into 2.6.15, and built it, but the module package wouldn't install because, unlike the source, it depended on a newer ndiswrapper-utils. So once again I booted back into 2.6.8 to download and install the newer ndiswrapper-utils. Which of course meant I had to remove my 2.6.8 ndiswrapper modules. Which meant that 2.6.8 wouldn't have wireless anymore. So then I booted back into 2.6.15, installed the ndiswrapper modules, tried to load them... and they wouldn't because they were built with gcc 3.x. I'd forgotten to install gcc-4.0 and apparently this didn't happen automatically. Which left me without wireless network on either kernel and about an hour wasted, and having to go looking for a wired network. > People upgrading from stable to testing shouldn't automatically get > their kernel upgraded unless they're using some of the kernel-latest > metapackages, and even then will have the older kernel installed. So, > I'm making a trade-off: > - reboot into an older kernel or grab ndiswrapper-source from unstable > and rebuild > vs. > - upgrade ndiswrapper-utils and not be able to use ndiswrapper w/ the > old *or* new kernel without rebuilding (and there's no guarantee the > rebuild will work). Well, it seems that I seem to have managed to hit every one of the possible failure modes here. Maybe unlike real debian users I don't know what I'm doing. -- see shy jo
signature.asc
Description: Digital signature
