[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: how to increase space for tmpfs /tmp



On 20120331_003333, Andrei POPESCU wrote:
> On Vi, 30 mar 12, 14:44:37, Paul E Condon wrote:
> > 
> > No. You misunderstand me. There is a new extra requirement on TMPDIR, a
> > restriction on ones choise of its value. A directory entry on a disk
> > file system is not enough. It must be a directory entry that has a line
> > in /etc/fstab that enables its use as a mount point to real separate
> > partition. At least that is the way it is now. If this restriction were
> > removed by some change in the implementation that I know not how to do...
> > then your suggestion would likely work and the old way of using /tmp
> > would also work. 
> 
> I really have no idea what you are talking about, but there is no new
> requirement on TMPDIR.

Perhaps you are thinking of an officially published requirement. What
I mean 'requirement' is a condition which by trial is found to be
necessary for it to work.  I tried cases that did not meet this
condition and none of them worked, and conditions which did meet this
condition did work. The test was simple.  I found a file of output
from find on a large file structure and tried to sort in on inode
number. I got an error message about insufficient file space without
any attempt at a fix, then I tried several values for TMPDIR and
several different entries in /etc/fstab. Some combinations
successfully sorted the file. Some aborted with the same error message
as the base trial.  The pattern was clear. There must be an entry in
/etc/fstab and the entry must be usable to mount an existing partition
on an existing disk in the plain-vanilla traditional way. Putting an
entry in that would choke on mount -a, also did no good in letting
sort run to completion.

But Andrei, I don't think either of us is going to actually fix the
problem. And I do think that we have already generated enough noise
here to attract the attention of someone who can. That person, or
persons, has already probably tried to verify my claim, and has
already succeeded, or not. The process is underway. Let us stop this
fruitless back and forth and wait and see. I have recieved two
messages telling me that the problem is surely not in coreutils, which
I do not doubt. But no one yet has told me that they have tried to
repeat my test and have not found a problem. But maybe I have
misunderstood some messages. In which case it is still worthwhile
stopping this discussion, so let us do so.

Peace.
>   
> > In UNIX all directories are files ... special files that serve a
> > special system defined function, but files in the sense that they are
> > not inodes, or sectors, or blocks, etc. Linux follows UNIX on this
> > innovation of long ago.
> 
> From this point of view, sure, but this is not what I was talking about.
> 
> > > Err... your original /tmp is a directory on / not a file[1] and if you 
> > > don't mount anything there your system will happily use the available 
> > > space on / (the root partition).
> > > 
> > > [1] unless you had a dedicated partition, but AFAIK in such a case you 
> > > wouldn't get a tmpfs anyway
> > 
> > I don't know why I get a tmpfs. I didn't ask for it. I have supposed
> > it came with a new way of doing file handling in the system software,
> > part of a new implementation that was supposed to be a work-alike
> > replacement of the previous version.
>  
> /tmp on tmpfs has been optional before, it's just that the initscript 
> maintainers decided to make it default.
> 
> > I never had a dedicated partion for /tmp and now it is required. That,
> > to me, is a change. I fixed it when I learned that it is now required,
> > and I think it would be nice to go back to the old way because the old
> > way did not require a separate partition. But I repeat myself. Enough.
> > What happens will happen.
> 
> There is no *requirement* for /tmp to be a separate partition. I really 
> don't understand how you came to this conclusion.
>  
> > > P.S. I accidentally did some re-wrapping, how long do you set your 
> > > lines?
> > The default in mutt, whatever that is. I like defaults. That is the
> > main thing that originally attracted me to Debian. It offered defaults
> > that worked. 
> 
> Mutt uses an external editor for writing e-mails
> 
> ,----[ man muttrc ]
> | editor
> |     Type: path
> |     Default: “”
> |
> |     This variable specifies which editor is used by mutt. It defaults to the 
> |     value of the $VISUAL, or $EDITOR, environment  variable, or to the 
> |     string “/usr/bin/editor” if neither of those are set.
> `----
> 
> Kind regards,
> Andrei
> -- 
> Offtopic discussions among Debian users and developers:
> http://lists.alioth.debian.org/mailman/listinfo/d-community-offtopic

OK. I don't know how to fix it, and am at a loss to find out how. Lines
do get longer as they are quoted. I use emacs to write mail.

-- 
Paul E Condon           
pecondon@mesanetworks.net


Reply to: