Re: CVS commit: src/sys

[christos%zoulas.com@localhost (Christos Zoulas)]

On Jan 11, 11:26am, ad%netbsd.org@localhost (Andrew Doran) wrote:
-- Subject: Re: CVS commit: src/sys

| Hi,
| 
| On Sun, Jan 11, 2009 at 02:45:56AM +0000, Christos Zoulas wrote:
| 
| > Log Message:
| > merge christos-time_t
| 
| A few things I have noticed so far:
| 
| - adjtime1() leaks timecounter_lock.
| 
| - #ifdef LFS doesn't belong under kern/ or in the compat module. The shim
|   should live in the LFS source. I realize that LFS syscalls are not
|   currently installed - this is a bug.
| 
| - The compat module now needs to be marked as depending on the aio module,
|   unless the compat shim lives in sys_aio.c. I'd tend toward putting it in
|   the compat mod but it could also live with aio. I'm suggesting the
|   opposite of the LFS case here since it's a basic kernel facility without
|   a big mess of depends and 100's kb of code behind it, unlike LFS.
| 
| - compat_mod.c needs an ifdef AIO for the case where we build a monolithic
|   kernel without AIO (and a -DAIO in the module makefile so it gets picked
|   up when building that way).
| 
| There is one example of ugly depends handling for the MODULE macro in
| linux_mod.c (it's there to cater for monolithic builds and differerences
| between architectures, because e.g. a.out doesn't make sense on NetBSD/foo).

There is also the NTP syscall issue that I am not sure how to deal with.
I think that you understand and know how to fix those problems better than
I, since you encountered them before. I could try to fix them, but I imagine
that it would be a lot easier for you.

christos