[lkml]   [1997]   [Aug]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: NFS bug in 2.1.48
Systemkennung Linux <> writes:

> > > #!/bin/bash
> > > cat >/tmp/xxx <<BUGGY
> > > asdf
> > > BUGGY
> [...]
> > Your little script, however, does NOT result in this error (talking
> > about 2.1.49). But there definitely is something wrong. What really
> > strikes me that both constructs don't fiddle with file descriptors
> > located on a NFS file system, only with pipes between processes. That
> > -ESTALE error _should_ be impossible.
> Here documents are placed in a file in $TMPDIR by the shell. They do
> not use pipes in any shell I know of.

Oh, sorry. Didn't know that. But then, what do you think of the
following little shell script:

sh -s << _HERE_
ls -la $TMPDIR

If the Here document is placed in a file in $TMPDIR, then it should
show up in the output of `ls -la'. But it doesn't. I really don't
think that a decent shell implementation should create another file in
$TMPDIR for Here documents. And I'm pretty sure that at least bash
feds it to the standard input of the command using a pipe.

However, I didn't exmine the source code of bash or ksh or tcsh.
You may still be right, but I don't think so.

But all of the above doesn't touch the real problem: that bogus ESTALE
error (in the case of tclsh I _AM_ sure that no tmp file is created by
tclsh for reading the stderr of "exec some_command", tclsh uses pipes)



 \ /
  Last update: 2005-03-22 13:40    [from the cache]
©2003-2014 Jasper Spaans. hosted at Digital Ocean