keithe@sail.LABS.TEK.COM (Keith Ericson) (03/24/91)
Has anyone else experimented with ESIX's BSD filesystem - for /usr or anwhere else? (I have a beta copy of Rev. D.; I assume it's up to production code.) I lucked (?) into the availability of the BSD filesystem for ESIX when I was having trouble installing onto an ESDI drive (Maxtor XT-4380E; Adaptec 2322B-8 controller). In my many attempts to make the usr file system by hand I used /etc/ffs/mkfs instead of /etc/ffs/newfs; during the file creation process it informed me that it was creating a BSD file system with 255 character filenames - a good thing if you ask me. In attempting to find out more about this ability to create BSD filesystems I ran strings on /etc/mkfs and discovered, I think, that one _should_ be able to specify the BSD file system during the normal system installation if the "What kind of a file system do you want: 1=normal 2=FFS" would accept "0" as a valid respones. Instead, "0" is rejected as invalid during normal system installation. Anyway, I got /usr created as a BSD file system. (One thing I haven't found out yet is just _where_ the description of the filesystem type lives - in the boot block somewhere?) Now I can create files names with names like "this_is_a_very_long_file_name" and they actually retain all the characters. HOWEVER (you saw this coming, right) not everything works like it used to. The 'ls' command, by itself, spits out the directory contents. But piping the output of ls to anything (ls | cat) creates filenames truncated to 11 characters. Use of wild cards is broken: $ touch file1 $ touch file2 $ touch file1a $ touch file1b $ ls file1* file not found Piping ls into cpio results in cpio having no filenames handed to it. Weird? Weird. Anyone know what's going on? Should I go back to FFS? (help *) KEITHE()
bob@rancor.UUCP (Bob Willcox) (03/25/91)
In article <9194@sail.LABS.TEK.COM> keithe@sail.LABS.TEK.COM (Keith Ericson) writes: > ... questions about use of ESIX's FFS filesystem ... I have been using ESIX 5.3.2 Rev D on obiwan since last August with an Ultrastor ULTRA 12(F) ESDI controller and two Maxtor 4380E's with virtually no problems. Note that this is for *all* of the filesystems on the system. With the help of a patch (32BETA2) from ESIX's bbs and articles posted by Chin Fang (fangchin@elaine41.stanford.edu) and Mike Burg (mburg@unix386.Convergent.COM) I have now converted all but my root filesystem to use long file names (default from ESIX is for System V directory format). So far (still early) everything seems to work as expected. I have not noticed any truncation of filenames by the standard system utilities as you described (the patch does not change any of these, it only replaces the /etc/conf/pack.d/ffs/* files). I did have to rebuild rcs and compress with some (minor) modifications so that they no longer thought filename lengths were limited to 14 characters. Also, some of the GNU fileutils that I had previously built were broken (e.g., ls) so I rebuilt them as well. One of the changes that I made was to /etc/ffs/newfs to remove its use of the -S flag when invoking /etc/ffs/mkfs (as per Chin Fang's comments). Now whenever I make a ffs filesystem with newfs it makes it with the BSD directory format (which, of course, is what I want so long as it works). I did this on both my harddisk and boot floppys. > >HOWEVER (you saw this coming, right) not everything works like it used to. >The 'ls' command, by itself, spits out the directory contents. But piping the >output of ls to anything (ls | cat) creates filenames truncated to 11 >characters. Use of wild cards is broken: > > $ touch file1 > $ touch file2 > $ touch file1a > $ touch file1b > $ ls file1* > file not found > >Piping ls into cpio results in cpio having no filenames handed to it. All of this works correctly on my system. > >Anyone know what's going on? Should I go back to FFS? My experience so far has been positive...your mileage may vary. > >(help *) KEITHE() -- Bob Willcox ...!{rutgers|ames}!cs.utexas.edu!romp!rancor!bob Phone: 512 258-4224