The afs-script makes several changes on the system and in the configuration files. In my opinion the configuration should be clear before the script starts. At first the script has only the job to read the config-files. Then check the parts which are needed for starting processes. If parsing the config-files failes then the start script should abort with a warning or a message for solution. No reconfiguration should be done by the start-script. Example: -afs can handle cache in several modes: --mounted partition (ext2 and ext3) --directory --memory The actual afs-script checks only the mount-part with ext2 partition and creates the cacheinfo-config-file if it not exists. The cache-size is calculated automatically. The user gets no information about this arbitrary action. This looks like operating system "Windows". Lets make the CellServDB, ThisCell, afs.conf and cacheinfo manually or give us an example-file. We will check and modify it. If you want to make a Windows-Like sytem give us a GUI for graphical configuration.
What's the status here? This bug has been untouched for over two years... :/
Er, I assume he means openafs? If so the wrong people are cc'd CC'ing current maintainers unless I'm horribly wrong, in which case please correct me ;)
(In reply to comment #0) > The afs-script makes several changes on the system and in the configuration > files. In my opinion the configuration should be clear before the script starts. I have always agreed with this, from the moment I started revamping the openafs init-scripts :) It is true that the old scripts did some weird things, like creating the mentioned cacheinfo file, but this is currently done by the ebuild. The user can change the defaults by editing the cacheinfo file, it is never overwritten. > The actual afs-script checks only the mount-part with ext2 partition and creates > the cacheinfo-config-file if it not exists. The cache-size is calculated > automatically. The user gets no information about this arbitrary action. This > looks like operating system "Windows". Lets make the CellServDB, ThisCell, > afs.conf and cacheinfo manually or give us an example-file. We will check and > modify it. If you want to make a Windows-Like sytem give us a GUI for graphical > configuration. The checking of the "mount-part with ext2 partition" has been eliminated from the script. In a more unix-like manner now, we assume the user has chosen sane settings and that the right partition is mounted at the given cache-dir. This gives you the option of having the cache on your root partition now, which is the result of a default installation as it is. The kernel-module naming is also done by the ebuild. After installation, there is just a single openafs-module with a fixed name, again eliminating a big part of the init script. In short, this bug report is obsolete by now because of completely new init-scripts, and I believe I have addressed the issues mentioned. Unless someone disagrees, this can be closed as fixed.
> In short, this bug report is obsolete by now because of completely new > init-scripts, and I believe I have addressed the issues mentioned. Unless > someone disagrees, this can be closed as fixed. > I was just sifting through maintainer-needed, if it's fixed, worksforme ;0
