Re: /dev/null misconfiguration on caballero?
[no response since Monday, adding ia64@buildd.debian.org]
On Mon, Jul 16, 2012 at 04:54:01PM -0700, Ben Pfaff wrote:
> Looking at:
> https://buildd.debian.org/status/fetch.php?pkg=openvswitch&arch=ia64&ver=1.4.2%2Bgit20120612-4&stamp=1342474173
> and scrolling down, I see:
>
> ## ---------------------- ##
> ## Detailed failed tests. ##
> ## ---------------------- ##
>
> # -*- compilation -*-
> 324. json.at:329: testing input may not be empty ...
> ../../tests/json.at:331: test-json /dev/null
> --- - 2012-07-16 21:26:10.144150241 +0000
> +++ /build/buildd-openvswitch_1.4.2+git20120612-4-ia64-gP4sNv/openvswitch-1.4.2+git20120612/_debian/tests/testsuite.dir/at-groups/324/stdout 2012-07-16 21:26:10.000000000 +0000
> @@ -1,2 +1,2 @@
> -error: line 0, column 0, byte 0: empty input stream
> +error: line 1, column 0, byte 3: syntax error at beginning of input
>
> 324. json.at:329: 324. input may not be empty (json.at:329): FAILED (json.at:331)
>
> This implies that /dev/null isn't empty, because the "test-json"
> program was able to read at least 3 bytes from it. That can easily
> happen in a chroot if someone forgets to "mknod /dev/null c 1 3" and
> later someone writes data to /dev/null as root. Any chance this
> happened?
>
> (I'd be more inclined to blame the upload itself except that -3 built
> successfully everywhere, -4 is building successfully everywhere other
> than ia64, and the only change between -3 and -4 was deleting a couple
> of lines from a file not used in the unit tests.)
>
> Thanks,
>
> Ben.
Reply to: