Re: 5.0-DP2 questions
On Fri, Nov 29, 2002 at 12:11:42PM -0500, Robert Ames wrote: 2. My machine is a Pentium 166 with only 16 MB of RAM. I'm trying to rebuild the kernel and so far the compile has been running for almost 24 hours and it's not finished yet. Is this to be expected? Yes. gcc 3.x is slower, and the kernel contains more code. Your machine is probably swapping a lot just doing the compilation, which will make it even slower. 3. When trying to rcp files to this machine I get a rshd: Login incorrect error. I have inetd configured and running and a .rhosts file in place (with proper permissions). I'm assuming this might be PAM related. Any suggestions? Can you log in with plain rsh? Do the manual pages or release notes describe any relevant changes? Kris msg47747/pgp0.pgp Description: PGP signature
Re: 5.0-DP2 questions
On Friday 29 November 2002 12:12 pm, Kris Kennaway wrote: On Fri, Nov 29, 2002 at 12:11:42PM -0500, Robert Ames wrote: 2. My machine is a Pentium 166 with only 16 MB of RAM. I'm trying to rebuild the kernel and so far the compile has been running for almost 24 hours and it's not finished yet. Is this to be expected? Yes. gcc 3.x is slower, and the kernel contains more code. Your machine is probably swapping a lot just doing the compilation, which will make it even slower. Out of curiosity, how much slower is a 5.x kernel compilation than a 4.x, on average? My 486, 66 MHz and 16 MB RAM, compiles a 4.x kernel in about 3 hours. Thus by Robert's data point, -current seems at least 10-15 times slower... -David -- On the whole I am against mass murder. I rarely commit it myself, and often find myself quite out of sympathy with those who make a habit of it. -Bernard Levin Astronomy and Astrophysics Center The University of Chicago To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: 5.0-DP2 questions
Kris Kennaway wrote: 3. When trying to rcp files to this machine I get a rshd: Login incorrect error. I have inetd configured and running and a .rhosts file in place (with proper permissions). I'm assuming this might be PAM related. Any suggestions? Can you log in with plain rsh? Do the manual pages or release notes describe any relevant changes? Yes, I can log in with plain rsh. And no, I didn't notice any relevant changes in the documentation. _ Tired of spam? Get advanced junk mail protection with MSN 8. http://join.msn.com/?page=features/junkmail To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: 5.0-DP2 questions
From: David Syphers [EMAIL PROTECTED] Reply-To: [EMAIL PROTECTED] Date: Fri, 29 Nov 2002 13:04:47 -0600 [Well, I'm Cc:ing -current anyway -- dhw] Out of curiosity, how much slower is a 5.x kernel compilation than a 4.x, on average? My 486, 66 MHz and 16 MB RAM, compiles a 4.x kernel in about 3 hours. Thus by Robert's data point, -current seems at least 10-15 times slower... OK; in each of the following, I was building the kernel as part of the process of upgrading from yesterday's -STABLE or -CURRENT, respectively. In each case, we are comparing -STABLE and -CURRENT running on the *same* hardware -- not merely configured similarly; each machine is set up to multi-boot, and runs -STABLE on slice 1 and -CURRENT on a different slice (3 for the laptop; 4 for the build machine). I track each of -STABLE and -CURRENT on a daily basis on each machine. First, the laptop: g1-9(4.7-S)[1] grep '^ Kernel' current stable-1 current: Kernel build for LAPTOP_30W started on Fri Nov 29 08:59:55 PST 2002 current: Kernel build for LAPTOP_30W completed on Fri Nov 29 09:35:30 PST 2002 stable-1: Kernel build for LAPTOP_30W started on Fri Nov 29 06:12:25 PST 2002 stable-1: Kernel build for LAPTOP_30W completed on Fri Nov 29 06:22:09 PST 2002 g1-9(4.7-S)[2] And now, the build machine: freebeast(4.7-S)[1] grep '^ Kernel' current stable-1 current: Kernel build for FREEBEAST started on Fri Nov 29 06:59:37 PST 2002 current: Kernel build for FREEBEAST completed on Fri Nov 29 07:25:28 PST 2002 stable-1: Kernel build for FREEBEAST started on Fri Nov 29 05:14:10 PST 2002 stable-1: Kernel build for FREEBEAST completed on Fri Nov 29 05:21:02 PST 2002 freebeast(4.7-S)[2] So: -STABLE -CURRENT Laptop 09:4435:35 Build machine 06:5225:51 I don't use -j for building kernels; I expect that the 2nd CPU on the build machine isn't all that significant for this workload. On the other hand, the slower disk drive in the laptop is likely fairly significant. In each case, the -CURRENT kernel that is running ( the one that is being built) has WITNESS , INVARIANTS, and DIAGNOSTIC defined. The laptop is a 750 MHz PIII with 256 MB RAM; the build machine is a 2x876 MHz PIII with 512 MB RAM. That should, at least, provide a reasonably valid set of comparisons. Cheers, david (links to my resume at http://www.catwhisker.org/~david) -- David H. Wolfskill [EMAIL PROTECTED] I have no confidence in results obtained through the use of Microsoft products. To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: 5.0-DP2 questions
On Friday 29 November 2002 01:32 pm, David Wolfskill wrote: ... That should, at least, provide a reasonably valid set of comparisons. Thanks. I suppose Robert's results might be abnormally long if -current requires a lot more memory than -stable, thus requiring a lot of swap, as Kris pointed out. Looks like my 486 won't be jumping to -current soon :) -David -- On the whole I am against mass murder. I rarely commit it myself, and often find myself quite out of sympathy with those who make a habit of it. -Bernard Levin Astronomy and Astrophysics Center The University of Chicago To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: 5.0-DP2 questions
On Fri, Nov 29, 2002 at 02:02:42PM -0600, David Syphers wrote: On Friday 29 November 2002 01:32 pm, David Wolfskill wrote: ... That should, at least, provide a reasonably valid set of comparisons. Thanks. I suppose Robert's results might be abnormally long if -current requires a lot more memory than -stable, thus requiring a lot of swap, as Kris pointed out. Looks like my 486 won't be jumping to -current soon :) It's often more efficient to use binary installations/upgrades than source, on slow machines. For example, I build world on a fast machine, mount via NFS and then installworld on my slower machines. Kris msg47755/pgp0.pgp Description: PGP signature
Re: 5.0-DP2 questions
On Fri, Nov 29, 2002 at 02:28:35PM -0500, Robert Ames wrote: Kris Kennaway wrote: 3. When trying to rcp files to this machine I get a rshd: Login incorrect error. I have inetd configured and running and a .rhosts file in place (with proper permissions). I'm assuming this might be PAM related. Any suggestions? Can you log in with plain rsh? Do the manual pages or release notes describe any relevant changes? Yes, I can log in with plain rsh. And no, I didn't notice any relevant changes in the documentation. Thanks, that's a useful data point for someone who can investigate this further. Kris msg47756/pgp0.pgp Description: PGP signature
Re: 5.0-DP2 questions
Date: Fri, 29 Nov 2002 12:20:38 -0800 From: Kris Kennaway [EMAIL PROTECTED] It's often more efficient to use binary installations/upgrades than source, on slow machines. For example, I build world on a fast machine, mount via NFS and then installworld on my slower machines. Quite so -- not only more efficient, but also less painful. :-} Indeed, that is how my build machine achieved that designation: I install -STABLE snapshots built on it about every 2 weeks or so onto my firewall a macihne that acts as the externally-visible Web server. (And it would be faster less hassle for me to treat my laptop similarly; on the other hand, I wanted to be able to compare UP vs. SMP if Something Weird(tm) were to happen. I also wanted to be sure that I had an independent complete (and portable) build environment on my laptop -- complete with its own copy of the FreeBSD CVS repo.) Cheers, david (links to my resume at http://www.catwhisker.org/~david) -- David H. Wolfskill [EMAIL PROTECTED] I have no confidence in results obtained through the use of Microsoft products. To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: 5.0-DP2 questions
On Fri, 29 Nov 2002, David Syphers wrote: On Friday 29 November 2002 12:12 pm, Kris Kennaway wrote: On Fri, Nov 29, 2002 at 12:11:42PM -0500, Robert Ames wrote: 2. My machine is a Pentium 166 with only 16 MB of RAM. I'm trying to rebuild the kernel and so far the compile has been running for almost 24 hours and it's not finished yet. Is this to be expected? Yes. gcc 3.x is slower, and the kernel contains more code. Your machine is probably swapping a lot just doing the compilation, which will make it even slower. Out of curiosity, how much slower is a 5.x kernel compilation than a 4.x, on average? I'm not sure about 4.x, but a -current kernel with no modules takes about 3 times as long as a RELENG_3 kernel compiled by the 4.x compiler used to take (about 130 seconds instead of 43 seconds on an Athlon 1600 overclocked. The kernels are supposed to have a similar set of options. All times are all times are after running make depend which takes about 8 seconds for RELENG_3 and 11 seconds for -current. gcc-3 in April 2002 pessimized the compile times from 76 seconds to 114 seconds for -current and from 43 seconds to 66 seconds for RELENG_3. Further development of -current pessimized the compile time from 114 seconds to 130 seconds. Compiling LINT took 437 seconds on Sep 22. IIRC, compiling modules takes about the same time as compiling LINT. My 486, 66 MHz and 16 MB RAM, compiles a 4.x kernel in about 3 hours. Thus by Robert's data point, -current seems at least 10-15 times slower... Ouch. I remember being happy when upgrading from a 486/33 with 16MB to a 486DX2/66 with 32MB reduced my kernel compile time from about 16 minutes to about 9 minutes. Your 16MB of RAM is probably not nearly enough for today's bloat. Look at the real, user and system times and systat/vmstat/top to see if there is a lot of idle time caused by waiting for disks and/or paging to disk. Bruce To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
