update-initramfs (latest update) always fails during kernel install

[David A. Cobb]

ONEIRIC XUbuntu

During a large batch of upgrades, I found a persistent failure to do the kernel update. The error message stated that the post-install script returned an error 127 from update-initramfs, at or near line 1010.

Investigating the script, I find the only command in the vicinity is a "rm -f ${initramfs}.new" .
But the problem appears to be that the return code ( $? ) is tested BEFORE calling the command.

The "diff -U6" output below shows both the way it was, and the way (IMNSHO) it should be. If the $? was intended to capture the result of some previous command, I would strongly suggest capturing it immediately after that command.

<code language="patch">
@@ -170,15 +170,16 @@
   OPTS="-v ${OPTS}"
  fi
  if mkinitramfs ${OPTS} "${initramfs}.new" "${version}"; then
   mv -f "${initramfs}.new" "${initramfs}"
   set_sha1
  else
- mkinitramfs_return="$?"
+#### <superbiskit> 2012-01-24: capture return AFTER remove
   remove_initramfs_bak
- rm -f "${initramfs}.new"
+ mkinitramfs_return="$?"
+ rm -fv "${initramfs}.new"
   if [ "$mkinitramfs_return" = "2" ]; then
    # minversion wasn't met, exit 0
    exit 0
   fi
   echo "update-initramfs: failed for ${initramfs} with $mkinitramfs_return." >&2
   exit $mkinitramfs_return

//
</code>

[David A. Cobb (superbiskit)]

Well, that really really needs an edit!!

1) The line number is 177;

2) The command was the function remove_initrdfs_bak.

THUS:
OLD=
               OPTS="-v ${OPTS}"
        fi
        if mkinitramfs ${OPTS} "${initramfs}.new" "${version}"; then
                mv -f "${initramfs}.new" "${initramfs}"
                set_sha1
        else
                mkinitramfs_return="$?" ### <<< This is the problem
                remove_initramfs_bak ### Assuming this is the return code you wanted to capture
                rm -f "${initramfs}.new"
                if [ "$mkinitramfs_return" = "2" ]; then
                        # minversion wasn't met, exit 0
                        exit 0
                fi
                echo "update-initramfs: failed for ${initramfs} with $mkinitram\
fs_return." >&2
                exit $mkinitramfs_return
        fi
}

NEW=
       if mkinitramfs ${OPTS} "${initramfs}.new" "${version}"; then
                mv -f "${initramfs}.new" "${initramfs}"
                set_sha1
        else
#### <superbiskit> 2012-01-24: capture return AFTER remove
                remove_initramfs_bak
                mkinitramfs_return="$?"
                rm -fv "${initramfs}.new"
                if [ "$mkinitramfs_return" = "2" ]; then
                        # minversion wasn't met, exit 0
                        exit 0
                fi
                echo "update-initramfs: failed for ${initramfs} with $mkinitram\
fs_return." >&2
                exit $mkinitramfs_return
        fi
}

[David A. Cobb (superbiskit)]

Anyway, while it is regrettable if a "initramfs_bak" is left hanging around, that deserves nothing worse than a warning. There is still a good and valid initramfs file created. An error means (IMNSHO) the program was not able to perform its task.

[David A. Cobb (superbiskit)]

Bleach! My own damn fault. (There was an RCS directory where update picked it up).
OK to invalidate this bug.

[dino99 (9d9)]

This version has expired