[Sbcl-devel] Re: SBCL from CVS on FreeBSD

[Andreas F. <as...@bo...>]

Today, Walter C. Pelissero <wa...@pe...> wrote:
> Just compiled SBCL from CVS on FreeBSD, and it seems to make
> everything beside the Posix extensions:
[...]
> At the first glance, I'd say that at least in two cases everything
> seems to be alright.  It's just FreeBSD behaving rather peculiar
> (different errno).

Yeah. The mkdir.error.2 and rmdir.error.2 failures are bugs in the
freebsd kernel. I just sent a problem report to the FreeBSD bug
tracker regarding this behavior.

> Test SB-POSIX-TESTS::RMDIR.ERROR.5 failed
> Expected value: 13
> Actual value: #<TYPE-ERROR {4837B2B9}>.
>
> Test SB-POSIX-TESTS::STAT.ERROR.2 failed
> Expected value: 13
> Actual value: #<TYPE-ERROR {485913B1}>.

The problem with these two lies in the use of a literal #xffffffff
with MKDIR instead of something that was aconstructed from s-*
constants.

> I've proceded to amend the first two tests as follows:
>
> (deftest mkdir.error.2
> (handler-case
> 	(sb-posix:mkdir "/" 0)
> (sb-posix:syscall-error (c)
> 	(sb-posix:syscall-errno c)))
> #+bsd #.sb-posix::eisdir
> #-bsd #.sb-posix::eexist)
>
> (deftest rmdir.error.3
> (handler-case
> 	(sb-posix:rmdir "/")
> (sb-posix:syscall-error (c)
> 	(sb-posix:syscall-errno c)))
> #+bsd #.sb-posix::eisdir
> #-bsd #.sb-posix::ebusy)

Thanks, I added these two. 

> and cutting off the other two with a #-freebsd conditional.  While
> not particularly beatiful at least it's better than what the
> official FreeBSD port does:

Right. Hm, I think I have fixed the type errors The Right Way instead
(replaced the #xff... with a constant LOGIORed from s-* permission
bits). This makes sb-posix pass all the tests on FreeBSD, so it is now
committed in 0.8.6.7.

Thanks for reporting this,
-- 
Andreas Fuchs, <as...@ac...>, as...@ja..., antifuchs