Bill White <[EMAIL PROTECTED]> wrote: > >> I am going to use the problem with vim as a test bed to learn > >> to debug Hurd programs. > > > >Building stuff on ext2 these days is a challenge. I recommend > >replacing /hurd/ext2fs and /hurd/ext2fs.static with Roland's verbose > >versions at http://www.ai.mit.edu/people/roland/hurd/ and sending him > >any assertion failure output that goes to the console. > I will do this this evening. I have intended to have my source code > all on NFS, and only have the Hurd-required stuff on the ext2 disk. > Presumably this would be pretty much safe from filesystem corruption. > If there was an easy way to do it, I would probably put the build > environment on an nfs-mounted disk as well. Building gcc is not that > hard.
I've got my home dir mounted rw over NFS, and I have no problem using it for pure source trees, but I've had to put objdir on a local partition for two reasons: (1)The Hurd NFS leaves .nfs* files in directories, causing rm -rf to fail. (2)dlopen of shared libraries on NFS fails, I guess because mmap is not implemented. > I don't even mount my bootstrap debian > partition on the Hurd when its running. FWIW, I've had no corruption with /linux mounted readonly under Hurd. > Problems like this ext2 corruption problem are why I have > spent the last three weeks getting a second machine up and running > before doing any hard work. :-) I bought a second monitor and replaced PLIP with Ethernet on account of the Hurd.
