Chmouel,
> it's normal if your ls is a ls --color, ls do before a file on each
> files to see the filetype and colorise it. your perl script do only
> a opendir/readdir, it's only two call to system.
Thats it, thanks .. I did 'unalias ls' and time dropped from 10-15 mins
to 2 seconds. Then I did 'alias ls="ls --color"' and it took long time
again.
>> Also, it seems slower to copy a file into a directory with many
>> files than into one with few. Is this well a known thing? I have
>> guesses to why, but that would be pure speculation; do anyone know?
>
>Could you bring some benchmarks with the time(2) command.
Ok .. I am at my home box now, 2.0.30, Pentium 200, RH 5.1, maybe it
makes no sense to do it then, but I can repeat tomorrow if you wish.
Anyhow, I get the same result here: I moved 1000 small files (they
were symbolic links) from directory with ~30,000 files into directory
"temp". Then I first copied these files into an empty "temp2" directory,
(biobase) /hd/9/temp> time cp -d * ../temp2
0.01user 1.93system 0:04.11elapsed 47%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (79major+10minor)pagefaults 0swaps
Now I copy these ~1000 files back where they came from,
(biobase) /hd/9/temp> time cp -d * ../chromat_dir
0.01user 65.70system 2:10.44elapsed 50%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (79major+10minor)pagefaults 0swaps
.. 30-fold increase .. maybe an inefficient way of updating directory
information .. (pure speculation)
>> Another observation .. I was in the middle of editing /etc/fstab
>> with Emacs when guy downstairs presses the power button, forgetting
>> to see if someone works on it, also forgetting to shut down right.
>> Then, he couldnt boot the machine. We were asked during the boot
>> to fix file system errors, ran fsck and pressed y a few times
>> (what else can one do?) and try again. Still no boot, because now
>> /etc/fstab was gone. I have not tried to reproduce this (pulling
>> the plug while editing a file isnt fun), and I have not yet applied
>> the Mandrake updates (kernel and initscripts).
>
>Do the updates and it should fix your problem.
>
>> Again, has this been seen by others?
>
>Oh yes :-((.
Allright, I am happy to hear this. Its kind of an unfortunate error
then, almost wonder if might be reason enough to update your iso's (?)
for the sake of those newcomers who overlook that applying updates is
a must .. maybe ..
Thanks for being so helpful ..
Niels L
