|
Kernel Traffic Latest | Archives | People | Topics |
Wine Latest | Archives | People | Topics |
GNUe Latest | Archives | People | Topics |
| Czech |
| Home | News | RSS Feeds | Mailing Lists | Authors Info | Mirrors | Stalled Traffic |
Do you pine for the nice days of minix-1.1, when men were men and wrote their own device drivers?
Are you without a nice project and just dying to cut your teeth on an OS you can try to modify for your needs?
Are you finding it frustrating when everything works on minix? No more all-nighters to get a nifty program working?
Then this post might be just for you :-)
-- Linus Torvalds, 1991
Table Of Contents
Introduction
I'm going to try to change the schedule of KC Debian Hurd so it comes out Monday nights instead of Thursday nights, so next issue will be out in about a week and a half.Mailing List Stats For This Week
We looked at 125 posts in 371K.
There were 27 different contributors. 17 posted more than once. 10 posted last week too.
The top posters of the week were:
1. Stay Away From The libnfslock Package
7 Jun 1999 (2 posts) Archive Link: "HORROR!: System in unstable state...."
People: Brent Fulgham, Marcus Brinkmann
Brent Fulgham tried installing the libnfslock package from the Debian Hurd archive, and it completely locked his system. Marcus Brinkmann thought the problem might be solved by deleting /etc/ld.so.preload, but he really wasn't sure. He'd built that package in the first place, and hadn't been aware that it could cause harm; for the moment, he recommended just not installing it.
2. Hurd And Linux Coexistence; Hurd Portability
7 Jun 1999 (3 posts) Archive Link: "Hurd (fwd)"
People: Roland McGrath, M.C. Vernon, Marcus Brinkmann
M.C. Vernon forwarded a messaged from Philippe Nantermod, in which Philippe wanted to have the Linux kernel and the Hurd kernel on the same partition, without changing his Linux configuration. He'd also wanted to know if he could put the Hurd on his Indy SGI station.
Marcus Brinkmann replied that putting the Hurd kernel on a Linux partition was possible, but it wouldn't give a usable system unless the rest of the Hurd's OS was on its own partition. I.e. the Hurd kernel is not a drop-in replacement for the Linux kernel. It was also possible to put the rest of the Hurd's OS on a Linux subdirectory or diskfile, but Marcus couldn't recommend it, since it was mostly untested and undocumented.
He added that no, the Hurd wouldn't run on the Indy SGI, but (so far) only ran on Intel [3,4,5]86. But Roland McGrath replied, "There is a mips port of mach4 (the predecessor to gnumach) and the hurd done by Kazumoto Kojima <kkojima@rr.iij4u.or.jp>. It has not been updated in a while, and has not been integrated into gnumach. But the code is there for the mips architecture if somebody wants to take it up."
3. Latest Hurd Version; Hurd On AMD
7 Jun 1999 - 9 Jun 1999 (4 posts) Archive Link: "Hurd releases"
People: Brent Fulgham, Marcus Brinkmann
Nuno Emanuel F. Carvalho wanted to know the current version of the Hurd. Marcus Brinkmann replied that it was 0.2, but that the development snapshot was much more current.
Nuno also wanted to know if an AMD K6 400MHz processor would be okay for the Hurd, and Brent Fulgham replied that he'd had success with an AMD K6-2 266MHz, and didn't see a problem with the K6-400.
4. Dead Inetd And Solution
7 Jun 1999 - 9 Jun 1999 (5 posts) Archive Link: "Open ports..."
People: Marcus Brinkmann, Mark Kettenis, Michael Bacarella
Michael Bacarella found that inetd wasn't running on his system, and connecting to various standard ports (ftp, telnet, finger) resulted in an immediate disconnect. Mark Kettenis pointed out that inetd should have been started at bootup. Michael replied that the init scripts did attempt to start inetd, but start-stop-daemon exited with a complaint about /proc not being mounted. Marcus Brinkmann and Ian Main said that his start-stop-daemon must be old (Ian had rewritten it), and Marcus recommended upgrading 'dpkg', which contained the later version.
5. New Buggy gnumach; cron, gs, jade, jed, wget, tetex Successes
8 Jun 1999 (4 posts) Archive Link: "new gnumach on alpha and more packages"
People: Roland McGrath, OKUJI Yoshinori, Marcus Brinkmann
Marcus Brinkmann reported a number of uploads. He'd uploaded a new gnumach, but warned that there was some header file difficulty that prevented it from compiling. He also uploaded cron 3.0pl1-50, and pointed out that
if [ -x /etc/init.d/cron ] ; then
/etc/init.d/cron start
fi
should be added to libexec.
He also uploaded gs 5.10-1 (though without much testing), jade 1.2.1-4, jed 0.98.7-14 (also containing xjed), and wget 1.5.3-1.1, all of which seemed to work for him. He also uploaded minicom 1.82-3, completely untested; and tetex-bin 0.9.981113-2, which worked, but isn't the latest version.
He also indicated that he might do some work on sysvinit in the near future.
Jaakko Niemi was thrilled about having 'jed', and asked if klogd or syslogd were available for the Hurd. Marcus replied that the Hurd currently uses its own syslogd from inetutils; and that there was no klogd, but that OKUJI Yoshinori had implemented kmesg (though Marcus didn't know the status). Roland McGrath replied, "The klogd from the Linux syslogd could be adapted. With the kmsg device and the same stream device translator needed for keyboards, etc, we will have a /dev/kmsg that behaves much like /proc/kmsg in Linux."
6. Successful Hurd Install; Installing A Bootloader; Mounting A CD; Removing A Translator; ncftp Success; Lists Of Good Packages; apt-get Saga
8 Jun 1999 - 9 Jun 1999 (9 posts) Archive Link: "Mounting and more"
Topics: Apt
People: Mark Kettenis, OKUJI Yoshinori, Brent Fulgham, Marcus Brinkmann
Per Lundberg reported Hurd installation success, but he had a couple problems. First, he wasn't able to install Grub as a bootloader on his hard drive. He suspected it was because his bootdisk had grub 0.4, but the package he installed was 0.5; Per wanted to know where to find a bootdisk with grub 0.5. OKUJI Yoshinori replied that he could make the bootdisk himself, with
$ dd if=[somewhere]/stage1 of=/dev/fd0 bs=512 count=1 $ dd if=[somewhere]/stage2 of=/dev/fd0 bs=512 seek=1
Per also had trouble mounting his CD. He'd been giving the incorrect command 'settrans /dev/hd1 /hurd/isofs /mnt'. Mark Kettenis pointed out the correct command, 'settrans /mnt /hurd/isofs /dev/hd1', adding, "Make sure the translator on /dev/hd1 is set properly (run MAKEDEV hd1)."
Per also asked how to remove a translator, and Marcus Brinkmann pointed out the command 'settrans -fg /mnt', in which -fg stood for "force goaway". Mark added some detail (slightly restructured for quoting):
'settrans /mnt' removes the passive translator on mnt (the passive translator is recorded in the filesystem, and specifies what program becomes the active translator when you access the node.)
'settrans -a /mnt' removes the active translator, but not the passive translator. This means that the next time you access /mnt, a fresh translator is started.. (the active translator is the program actually running on top of the node to provide its contents.)
Other interesting options are -g' (ask the active translator go away (nicely) when you set the passive translator) and -f' (force the active translator to go a way). See 'settrans --help' for more information.
Per also had had a little trouble with apt-get trying to download stuff from the wrong directory. Marcus was thrilled that apt-get had compiled at all, and said he'd try compiling and uploading it (see below for Per's later attempt).
Finally, Per had added that once he got his system set up, he might be interested in porting a few packages. He was thinking of starting with ncftp, if no one was doing it. Marcus hadn't, but thought Brent Fulgham might be working on it. Brent replied that he wasn't either, though it would be a good thing to have around.
Later, under the Subject: NCFTP package, Per reported total success with ncftp, and Marcus posted a few relevant (incomplete) lists, welcoming any corrections or additions.
Packages requiring no patching:
ae (962-23.1) an (0.93-1.1) bc (1.05a-4) bison (1.25.90-2) bzip2 (0.9.0c-2) cvs (1.10.4-1) db (2.4.14) diff (2.7-18) diffstat (1.27-1) ed (0.2-17) flex (2.5.4a-4) grep (2.3-2) groff (1.11a-7.1) gzip (1.2.4-29) inetutils (1.3.2-4) jed (0.98.7-14) less (332-4.1) libglib (1.2) libident (0.21-7) libpng (1.0.2b-0.1) lynx (2.8.1-7) make (3.77-6) mawk (1.3.3-2) minicom (1.82-3) mpack (1.5-5) ncftp pcre2 (2.05-1) pinfo procmail (3.10.7-7) sed (3.02-1) sharutils (4.2-9) svgalib-dummy (1.2.13-2) tar (1.12-7) wget (1.5.3-1.1)
Packages which require only a patch already in the bug tracking system:
cpio (2.4.2-24, #37164) debianutils (1.11, #39126) exim (2.11-3) file (3.26-1, #37705) fortune-mod (9708-10, #37091) jade (1.2.1-4, #38518) libpaperg (1.0.3-12.1, #39036) man-db (2.3.10-69f, #39039)
Later, under the Subject: APT, Per reported partial success with apt-get. It made it all the way to the configure phase, where it ran into problems with unimplemented Hurd infrastructures, in particular libpthreads. Mark replied that he was working on implementing that, but progress was slow, and he had to give priority to fixing Hurd bugs at the moment.
There was a bit more discussion, focusing mainly on swap and gdb upgrading, and the thread ended.
7. 'man' Not Yet Working Under The Hurd
8 Jun 1999 - 12 Jun 1999 (13 posts) Archive Link: "Broken libdb2?"
People: Marcus Brinkmann
Per Lundberg reported that the libdb2 package (required by 'man') seemed corrupt, and in the course of discussion it came out that ftp.debian.org didn't default to binary mode for FTP transfers. Jonas Oberg fixed that problem, but the libdb2 packages weren't ready anyway. Marcus Brinkmann reported that the sources compiled successfully, but that was it. The binaries themselves were still completely broken.
8. Porting Debugging Tools To The Hurd
8 Jun 1999 (3 posts) Archive Link: "strace/ltrace"
People: Roland McGrath, Brent Fulgham
Brent Fulgham wanted to implement Linux strace/ltrace features on the Hurd, since he thought it would be very useful in tracking down problems. But he was curious how much effort he'd have to put out. Roland McGrath said that ltrace should be fairly easy to port to the Hurd. He went on, "the rough equivalent to strace for the Hurd would be an RPC tracing program. This would indeed be very useful for debugging, and I think would also be an excellent tool for helping people understand the Hurd design. A few people have started working on this at different times, but I haven't heard anything from anybody about it recently."
Brent replied that he'd try to get ltrace and the GNU nana debugging package (which he maintained) working for the Hurd, and only then move on to the more difficult strace question. Then later, under the Subject: asm/unistd.h and asm/ptrace.h and the Hurd, Brent had been working on porting ltrace, but he was confused by differences between the Hurd and Linux. After some back-and-forth with Roland, it came out that the equivalent of Linux system calls were all implemented as normal libc functions under the Hurd. That was what Brent was looking for, and the thread ended.
9. Screen Almost Ready
8 Jun 1999 (1 post) Archive Link: "more about screen"
People: Marcus Brinkmann
Marcus Brinkmann announced that screen still wasn't working, though it was very close. He gave a patch and a lot of instructions for folks interested in helping. The reason he was putting so much effort into it was because it is a relatively quick way to have virtual consoles on Hurd.
10. Hurd Profiling
8 Jun 1999 - 9 Jun 1999 (2 posts) Archive Link: "Hurd Latencies -- Questions..."
People: Brent Fulgham, Marcus Brinkmann
Brent Fulgham noticed that the Hurd was still noticably slower than Linux on the same hardware, and asked what sort of profiling had been done to locate the bottlenecks. Marcus Brinkmann admitted he was no expert on the subject, but he didn't think any profiling had been done. He added that "Making sure profiling works" was still an item on the task list.
11. Temporary Sysvinit Alternative
9 Jun 1999 (1 post) Archive Link: "poor man's sysvinit, please try"
People: Marcus Brinkmann
Marcus Brinkmann posted a tar file of a sysvinit replacement. To get it working, folks would have to make /libexec/rc a symlink to /etc/init.d/rcS, and reboot. He added, "I tried to minimalize the differences between the Hurd and Linux, but some stuff had to be deleted/changed."
12. Slight Problem (And Solution) Recompiling GNUMach
9 Jun 1999 (6 posts) Archive Link: "Recompiling GNUMach"
People: Brent Fulgham, Marcus Brinkmann
Per Lundberg couldn't recompile gnumach because he didn't have 'mig', and he wanted to know what 'mig' was. Brent Fulgham replied, "MIG is the (M)ach (I)nterface (G)enerator. I'm no expert, but I kind of think of it like IDL files in CORBA. You create an interface file for a server or other system process using a special descriptive language (codified in the MIG programming manual), then you run it through MIG and it generates some glue code that marshalls/demarshalls the RPC calls that are used as the native message passing protocol for Mach-based systems." He added, "This is probably a gross simplification, but there it is."
Marcus Brinkmann added that the mig package was in incoming and would be on the archive soon.
13. Autoprobing Hardware Misidentification
9 Jun 1999 (3 posts) Archive Link: "NIC mishap"
People: Marcus Brinkmann, Michael Bacarella
Michael Bacarella upgraded from gnumach_1.1.92-1.deb to gnumach_1.1.92-2.deb and found that his NIC was being detected as a 3c59x instead of ne2k. Marcus Brinkmann replied, "The problem is that the order of autoprobing seems to be suboptimal. Maybe the set of included drivers is also suboptimal." He added, "Frankly, as long as we don't have way to load modules at run-time, there is no easy solution. What we can do is provide a default kernel that at leasts boots on all system and if you have problems with network cards etc you need to roll your own."
14. Installation Difficulties
9 Jun 1999 - 10 Jun 1999 (5 posts) Archive Link: "my experiences"
Topics: FS: ext2
People: OKUJI Yoshinori, Brian Gregor
Tim Newsham tried installing the Hurd, and had many little problems, mainly having to do with docs, scripts, and the non-Debian distribution he was installing from (Red Hat).
After a bit of help, he got the system almost booting. But it would hang, on
hd02: badd access: block=28 count=2
end_request: I/O error, dev 03:02 sector 28
Hurd Server bootstrap: ext2fs.static[hd0s7] exec
OKUJI Yoshinori replied that he'd been investigating that very problem. He speculated, "Perhaps this error will happen when a drive is >8GB. Marcus's drive is also >8GB. To fix the error, it is neccesary to examine the glue code and the driver code entirely."
Under the Subject: Installation & boot problems, Brian Gregor had also tried to install the Hurd from a Red Hat box. He had an easier time of it, though. After changing all occurrences of "==" to "=" in the install scripts and resizing his partitions to <1Gb, he reported total success. He added, "One note to Red Hat users - I am using a v5.2 installation, and dpkg wanted the start-stop daemon installed. I found a C version of this in the /usr/doc/SysVinit-2.74/ directory, which compiled fine with the inclusion of "#include <errno.h>" in the beginning."
15. X On The Hurd
10 Jun 1999 - 12 Jun 1999 (4 posts) Archive Link: "Unidentified subject!"
People: Marcus Brinkmann
Jake Bishop asked if X was working for the Hurd yet, and Marcus Brinkmann replied that certain parts of it were, and certain parts (notably the actual servers) were not. He gave a URL for existing .debs, and added, "Don't ask to help with the servers if you aren't a Hurd or GNU/Mach hacker (or wanna-be :)"
16. More Doc Format Discussion
13 Jun 1999 (1 post) Archive Link: "Re: no doc format discussion, please"
Continuing the discussion from KC Debian Hurd, issue 1, article 4, Guy W. Hulbert insisted that doc format discussion should be allowed on the documentation mailing list. He strongly disagreed with GNU's anti-man-page policies, and suggested XML as a good format. There was no reply.
Sharon And Joy
Kernel Traffic is grateful to be developed on a computer donated by Professor Greg Benson and Professor Allan Cruse in the Department of Computer Science at the University of San Francisco. This is the same department that invented FlashMob Computing. Kernel Traffic is hosted by the generous folks at kernel.org. All pages on this site are copyright their original authors, and distributed under the terms of the GNU General Public License version 2.0. |