ATF work

Alan Somers asomers at
Wed Apr 2 17:36:44 UTC 2014

On Wed, Apr 2, 2014 at 11:34 AM, Peter Holm <peter at> wrote:
> On Wed, Apr 02, 2014 at 10:31:44AM -0600, Alan Somers wrote:
>> On Wed, Apr 2, 2014 at 2:53 AM, Peter Holm <peter at> wrote:
>> > I have uploaded a snapshot of the ATF work done for EMC Isilon,
>> > based primarily on Garrett Cooper's work.
>> >
>> >
>> > +
>> > chmod +x /usr/src/lib/libc/tests/net/gen_ether_subr
>> Holy patch queue, Batman!  There's fifty thousand lines of code in here!
>> I'm not going to read every line, but here's what I noticed:
> The patch is just a snapshot at what's being worked on, where the
> goal was to get the maximum number of working test scenarios. As
> you may have noticed the broken ones has just been marked as "skip"
> / WIP for now.
> I'll go thru your comments and fix the issues pointed out. Thank you
> for doing this!
> I just started looking at rump. It seems some tests needs the user
> mode tools like for example rump.ifconfig from
> rumpkernel/netbsd-userspace-src Do you know if anybody has looked
> at those?

rump requires extensive kernel support.  It would be great if FreeBSD
had it, but it's not going to happen any time soon.

> As for the format of the patch file; I use it all to time for tests?
> (cd /usr; patch -p2 < ~pho/FreeBSD-HEAD-ATF-Isilon-20140402.diff)

It only works if your source tree is at /usr/src.  Mine isn't.  What
command are you using to produce the diff?

>> All of the files from NetBSD should have a $FreeBSD:$ tag alongside
>> the NetBSD tag.
>> lib/libc/tests/net/getaddrinfo/Makefile has some commented out code
>> that should be removed.
>> So does lib/libc/tests/net/getaddrinfo/
>> Could lib/libc/tests/net/gen_ether_subr be converted to an awk script
>> by replacing the first line with "#!/usr/bin/awk -f" and removing the
>> invocation to awk?  Then it would have the correct syntax highlighting
>> in editors.
>> lib/libc/tests/net/Makefile has one tested comment out with the
>> comment "test uses rump".  Would it be possible instead to leave the
>> test in the build, but put "require.progs rump_server" in the relevant
>> test cases' heads?
>> It would be more idiomatic for lib/libc/tests/tls_dso/Makefile to
>> define a TESTSDIR variable and use that in the definitions of LIBDIR
>> and SHLIBDIR.
>> lib/libc/tests/gen/Makefile has several tests commented out.  Several
>> of the comments are confusingly short (what does "XXX: F_MAXFD DNE"
>> mean)?  And the "#ATF_TESTS+=   posix_spawn" line should be deleted.
>> lib/libc/tests/gen/posix_spawn/Makefile contains commented out code.
>> lib/libc/tests/arch contains subdirectories that FreeBSD doesn't
>> support and never will, like Vax.  They should be deleted.
>> It's unfortunate that all of t_sha2 is disabled.  I don't understand
>> the comment, either.  Is it disabled just because FreeBSD doesn't have
>> sha384 functions?  We have both sha1 and sha256 in libc.
>> lib/libc/tests/termios/Makefile has WARNS=4 commented out.  Why is that?
>> Some Makefiles define "MAN= ".  It is more idiomatic to define "NOMAN=".
>> Check your whitespace in tests/sys/kern/kern_fork_test.c and
>> tests/sys/kern/user_test.c.  I see lines with leading spaces.
>> I tried to test your patch, but I was unable to apply it.  It looks
>> like you did "diff -dur src /usr/src" or something like that.  The
>> problem is that the two paths need different levels of directories to
>> be stripped.  Neither patch nor svn patch can deal with that.  The
>> result is that either the newly created files appear in the wrong
>> place, or the tool patches /usr/src/Makefile when it's supposed to
>> patch /usr/src/tests/Makefile and /usr/src/libc/Makefile.  Could you
>> please try to regenerate the patch?
>> -Alan
>> >
>> > - Peter
>> > _______________________________________________
>> > freebsd-testing at mailing list
>> >
>> > To unsubscribe, send any mail to "freebsd-testing-unsubscribe at"
> --
> Peter

More information about the freebsd-testing mailing list