Subject: Re: Progress meter for fsck, revisited
To: Jason Thorpe <>
From: Greywolf <>
List: tech-userlevel
Date: 01/13/2004 13:29:55
Thus spake Jason Thorpe ("JT> ") sometime Today...

JT> These patches are based on his, but have some changes:
JT> 	- If -P is specified, then the master fsck process disables
JT> parallelization
JT> 	  so that the meter output does not interfere with each other.
JT> 	- For preen, a single overall progress meter is used, rather than
JT> 	  one for each pass.

Don't those two kind of conflict, since -p == (parallel | preen)?

[i.e.  I have never seen fsck go parallel without the -p option.]

I would say that you might want to have -P work with parallelisation
such that it only uses one meter.

JT> 	- The code is sharable by fsck utilities other than fsck_ffs.

This is good...

JT> 	- The code can be compiled out using -DSMALL.

Hmm, what if one wants the progress meter but not the getpw* and err*

Incidentally, has anyone tried compiling fsck_ffs with -DSMALL?  Last
time I did it, it blew up horribly because the stuff in inode.c (IIRC)
which referenced getpw* and err* was not properly protected with the
appropriate #ifdefs.

22 Ways to Get Yourself Killed While Watching 'The Lord Of The Rings':

#18: Imitate what you think a conversation between Gollum, Dobby and Yoda
    would be like.