cvs commit: src/sys/alpha/alpha support.s src/sys/i386/i386 swtch.s src/sys/kern kern_shutdown.c src/sys/sys systm.h

[Jun Kuriyama]

At Tue, 17 Feb 2004 21:57:41 +0900,
kuriyama wrote:
> But I personally feel this should be back out after re-reading and we
> should tackle traceback() improvements.

I checked something around backtrace().  A db_print_backtrace() is
depending on DDB, but as Bruce said, DDB is dangerous for default
feature for -stable release.

From my (non kernel hacker) understanding, I'm not sure whether this
is possible and preferred approach or not.  How about this approach?

o Add DDB_BACKTRACE_ONLY option.
o Compile minimal code for db_print_backtrace() into the kernel (some
  #ifdef storm?) when this option is enabled without DDB.
o Call backtrace() if this option is enabled like as DDB_TRACE.
o Enable this option in GENERIC by default.


-- 
Jun Kuriyama <kuriyama at imgsrc.co.jp> // IMG SRC, Inc.
             <kuriyama at FreeBSD.org> // FreeBSD Project