(In reply to comment #0)
> While performing rescue jobs in a chroot (changing rootfs type):

This is not a regular operation, right?

> # mkinitrd --help
> /sbin/mkinitrd: line 1557: cannot create temp file for here document: No such file or directory
> 
> # set | grep T.*MP
> TMP=/tmp/.private/root
> TMPDIR=/tmp/.private/root
> # ls /tmp/.private/root
> ls: cannot access /tmp/.private/root: No such file or directory
> 
> Indeed.

There are many ways to fool yourself.
Try "env TMPDIR= chroot" next time.

(In reply to comment #1)
> > While performing rescue jobs in a chroot (changing rootfs type):
> This is not a regular operation, right?
It is one of operations when:
- the stress is usually above usual;
- weird diagnostics don't really help.

> > # mkinitrd --help
> > /sbin/mkinitrd: line 1557: cannot create temp file for here document: No such file or directory
> > ls: cannot access /tmp/.private/root: No such file or directory
> There are many ways to fool yourself.
> Try "env TMPDIR= chroot" next time.
I'll sure forget it 'till next time.

Rather wanted to draw your attention to the fact that corner cases like this do matter; is it unreasonable to request that directory existence[/writability] check be performed before setting TMP/TMPDIR?

Lets handle this issue in chroot(1).

Thanks in advance!

6.11-alt1-2-gc4aa447

Your advance has arrived :-)