Date: Sat, 10 Apr 1999 19:51:12 -0600 (MDT) From: "Kenneth D. Merry" <ken@plutotech.com> To: grog@lemis.com (Greg Lehey) Cc: ache@nagual.pp.ru, bde@zeta.org.au, ken@plutotech.com, hackers@freebsd.org, cvs-all@freebsd.org, cvs-committers@freebsd.org, ache@freebsd.org Subject: Re: cvs commit: src/usr.sbin/config main.c Message-ID: <199904110151.TAA34645@panzer.plutotech.com> In-Reply-To: <19990411103824.Y2142@lemis.com> from Greg Lehey at "Apr 11, 1999 10:38:24 am"
next in thread | previous in thread | raw e-mail | index | archive | help
Greg Lehey wrote... > On Saturday, 10 April 1999 at 18:22:25 +0400, Andrey A. Chernov wrote: > > On Sun, Apr 11, 1999 at 12:15:43AM +1000, Bruce Evans wrote: > >> The -s changes should be backed out completely. They were as uncouth > >> as making -l the default for ls and adding a deprecated -S flag to give > >> normal behaviour. > > > > I agree. Anyone who need -g kernel space/time bloat must use -g directly. > > Please ask for removing not me but '-s change' author. > > On Saturday, 10 April 1999 at 17:41:19 -0600, Kenneth D. Merry wrote: > > Bruce Evans wrote... > >>> Modified files: > >>> usr.sbin/config main.c > >>> Log: > >>> add -s to usage > >>> > >>> PR: 11056 > >>> Submitted by: Nickolay N. Dudorov <nnd@mail.nsk.ru> > >> > >> The -s changes should be backed out completely. They were as uncouth > >> as making -l the default for ls and adding a deprecated -S flag to give > >> normal behaviour. > > > > I agree. > > This was discussed at great length on -hackers before the changes were > made. People were (after a while) unanimously in favour of the > change. It would have been nice if you had responded there before I > did it, rather than afterwards. > > Bruce explained that he doesn't read -hackers. OK, in that case we > need somewhere else to discuss this kind of proposed change, rather > than making changes and then having people wanting them backed out. > Suggestions? I don't have any suggestions for that, but I will say that I tend to ignore most of the stuff on -hackers. It often tends to be more noise than signal. -current is often that way as well. The amount I read on either list varies with available time and interest in the subject. > In any case, it's all very well to "agree", but none of you have > presented any arguments for your viewpoint. In particular you, Ken, > have in the past asked people to use a debug kernel to get adequate > information after a panic. If no symbols is the default, 99% of > people who have problems won't be able to report them. As I proposed > in the discussion on -hackers, this solution should make everybody > happy: there is no difference in the running system, on modern > machines it only takes fractionally longer to build a debug kernel, > and it helps a lot in finding problems. Of course I've asked people to build debugging kernels. I just did yesterday, in fact. I've also been bitten by not having debugging symbols for a kernel on a machine that's paniced. > To summarize the changes: > > By default, config configures a kernel with full debug symbols. If > you really want to build one without symbols, you use the -s option. > > 'make install' always installs a kernel without debug symbols. If > you have a debug kernel (no -s option) and want to install it, use > the 'install.debug' target. The debug kernel is called kernel.debug > in the kernel build directory. The non-debug kernel is called > kernel, sa before, and is created by using objcopy --strip-debug, so > the code in the non-debug kernel is identical to the code in the > debug kernel. It appears that it's possible to have difference in > the code between kernels built with symbols and without, but this > issue does not arise here. > > If you still think the change to the default should be backed out, > that doesn't affect the changes to the Makefiles: if you have a debug > kernel, it was previously difficult to install a stripped version. > Now you have two targets which give you the choice. I really don't feel strongly about it either way. I almost always build kernels with debugging symbols in any case, but not everyone needs them. My main concern was with the fair amount of extra space it takes to build and store debugging and non-debugging kernels. The difference with a sample kernel config file here is 6MB for the non-debugging version versus 27MB for the debugging version. I dunno, maybe it is worth the extra disk space to get debugging symbols in case of a panic. You have made a reasonable case for it. In any case, consider my objection withdrawn. Ken -- Kenneth Merry ken@plutotech.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199904110151.TAA34645>