Hurd Traffic #87 For 20 Apr 2001

Editor: Zack Brown

By Paul Emsley

Mach 4 ( | Hurd Servers ( | Debian Hurd Home ( | Debian Hurd FAQ ( | debian-hurd List Archives ( | bug-hurd List Archives ( | help-hurd List Archive ( | Hurd Reference Manual ( | Hurd Installation Guide ( | Cross-Compiling GNUMach ( | Hurd Hardware Compatibility Guide (

Table Of Contents


Want to help write KC Debian Hurd? See the KC Authorship page (../author.html) the KC Debian Hurd homepage (index.html) , and the Thread Summary FAQ (../summaryfaq.html) . Send any questions to the KCDevel mailing list. (

Mailing List Stats For This Week

We looked at 397 posts in 2138K.

There were 84 different contributors. 38 posted more than once. 14 posted last week too.

The top posters of the week were:

1. Turle Info

27 Mar 2001 - 29 Mar 2001 (13 posts) Archive Link: "More turtle info... "

Summary By Paul Emsley

Topics: Apt

People: Jeff BaileyMarcus BrinkmannBen Collins

Jeff Bailey reported that his build machine was now producing results: He listed some packages that he would like to see fixed:

apt - 0.5.0 built fine, 0.5.2 and 0.5.3 did not.

db - This broke in the last version - I think their looking for profiles C libraries. Check the build log.

screen - This compiles, but doesn't work. Marcus said he may have some ideas on this one.

It seems as though db has been obsoleted by db2, producing a problem for turtle. Marcus Brinkman explained: "The problem is that there are now two source packages which produce the same binaries. How can an autobuilder know which of the two should be build?"

Ben Collins gave his explanation:

There was a hard reason. That reason being that I took over db2 and the build for it was so ugly I had to redo the package from scratch in order to sanely have a chance of fixing bugs. As this required a new full source upload, which dinstall doesn't allow without a new upstream version, I had to rename the source. Since db3 and db1 have specific source package names, I felt it was also a good idea to make db2 follow that anyway.

Now, James has assured me that the old db source will obsolete itself in time. The sparc autobuilder has not complained about this at all. Seems to me there is something specific about how hurd is handling this that makes it choke.

Marcus and Jeff discussed the communication of turtle and turtle1, Marcus finnishing with: "I would prefer to first work out the interface details. Basically we need a new state of a package ("claimed") and some infrastructure to claim/release packages and stuff in the data we get from the child machine. As I have to rework the state stuff anyway (for the error/retry stuff), it makes sense to defer this a bit. "

2. Gawk

30 Mar 2001 - 1 Apr 2001 (8 posts) Archive Link: "gawk :"

Summary By Paul Emsley

People: Jeff BaileyMarcus Brinkmann

Jeff Bailey found the reason why gawk won't compile is because make can't cope with ":" in the directory name.

There was some discussion on the horribleness of this problem, it's not really a Hurd issue. Marcus Brinkmann raised it on debian-policy and said that they should have a few days to consider. (It is Marcus' view that gawk should be fixed).

3. Sound, PPP, GGI

2 Apr 2001 - 3 Apr 2001 (5 posts) Archive Link: "Hurd queries"

Summary By Paul Emsley

Topics: FS: NFS, GGI

People: Marcus BrinkmannOgnyan KulevNiels MöllerBob HamGlenn Alexander

Bob Ham asked:

And in reply, Marcus Brinkmann wrote "The problem with sound is that those little sound devices are character rather than block, and we have nicer a glue layer in GNU Mach nor the libchannel abstraction yet. However, adding a simple glue layer for character devices (read: not terminals) is not so hard, and there are patches by Okuji and me floating around at, so I think it is possible to add OSS/lite and hack some simple translators (based on the stream translator we use for klog) to access them. "

There then follows some chat about linux modules, ALSA and Hurd smarts for sound.

In response to Ognyan Kulev's comment about PPP, Marcus replied: "I take the position that we have PPP support, but it is broken. The user space ppp software needs to be debugged. That's all. No new code to be written, no extensive porting. Just fixing the bugs. Anyone?"

Ognyan Kulev said: "GGI was mentioned in mailing lists a time ago" . Niels Möller clarified the issue wrt X: "I don't think the main reason for doing some KGI-like translator is not to replace X, it's just that it's Right Thing to do. If there's gfx hardware present, there ought to be a decent framebuffer-like device/translator for it, like for all other hardware. Having X bang the hardware directly is ugly."

Glenn Alexander had corresponeded with the Berlin dev list about Hurd compatability and got a response saying that Prague's Thread classes could be implemented with cthreads, there would be no need for pthreads. Also, the Hurd would need an ORB and a Console API.

Ognyan added: "I don't know about nfsd in /sbin. There is a NFS client which is a translator. Try /hurd/nfs --help. Its usage is something like that: settrans /mount/point /hurd/nfs But see --help output for more info."

4. Most Wanted Page

6 Apr 2001 - 11 Apr 2001 (10 posts) Archive Link: "Status pages?"

Summary By Paul Emsley

People: Jeff Bailey

Jeff Bailey asked if there was any value in keeping up a "most wanted" page for tasks that everyone can work on. There was some discussion about its value and Jeff decided to proceed, providing the link:

5. OSKit-Mach Booting

9 Apr 2001 - 12 Apr 2001 (10 posts) Archive Link: "Trouble with oskit-mach"

Summary By Paul Emsley

People: Bob HamRoland McGrath

Bob Ham said "I've been attempting to get oskit-mach up and running but get stuck at booting" , and provided some details.

Roland McGrath replied: "Anyone trying oskit-mach needs to first try some oskit example kernels. If the basic "hello" and "multiboot" example kernels work on your hardware, then move on to some example kernels that use the disk and ethernet. Only once you are happy with the generic oskit situation on your hardware should you start to worry about debugging oskit-mach itself." Roland went on to say "You seem to have presumed that this [i.e. failure to boot] has something to do with selecting devices, as it [often] does in gnumach. That is not the case for oskit-mach. You're going to have to get your hands dirty."

Bob said that he would and we haven't heard from him since.

6. Python On The Hurd

12 Apr 2001 (4 posts) Archive Link: "dupload: reading a conf' file in the local directory"

Summary By Paul Emsley

People: Jeff BaileyRoland McGrath

In reply to a response to his own question about dupload and python (presumably to the python mailing-list), Jeff Bailey pointed to a problem with using threads with python (being a Klingon, Jeff doesn't approve of building python without threads).

Jeff gave a description of his problem, saying "Hmm - There don't appear to be casts to (mutex_t). Does this actually work on other cthreads implementations, or is it deadwood? "

This prompted Roland McGrath to refer to an older thread, saying "I think the answer to your earlier question is that there are no other extant cthreads implementations that anyone is actually using." ("extant" seems to be Roland's Word-For-The-Month :), meaning "still in existance").

In the message Python && cthreads ( Jeff submitted a patch ( which got python to build. "D-Man" suggested 'make test' to use the python test suite.

7. SMP On The Hurd: Status Report

12 Apr 2001 - 13 Apr 2001 (5 posts) Archive Link: "SMP Hurd status report. Some questions, some comments."

Summary By Paul Emsley

People: Douglas HiltonRoland McGrathOgnyan Kulev

Douglas Hilton wrote "am using a custom microkernel now for Hurd. I was easily able to build oskit on Debian/Linux, and can run all the sample kernels. The oskit SMP sample kernel is detecting my two CPU'S, and seems to be working fine. " He went on to ask about (cross-)compiling oskit-mach and said that he had read that Roland was working on writing a book.

In reply, Roland McGrath said:

You can do whatever is easiest for you. You can build and install the oskit native, and then build oskit-mach using your native Linux compiler. That is probably the easiest thing. Or you can build the oskit native and then do "make install prefix=/usr/i386-gnu" (or whatever is appropriate) to install it where it will be found by your Hurd cross-compiler. Or you could cross-build the oskit for the Hurd (the oskit ends up the same either way). It really doesn't matter.

[WRT SMP gnumach:] As you've seen, the oskit SMP support works with current hardware (like yours), whereas gnumach doesn't even have any extant hardware support code.

[WRT the book:] Well, this is the first I've heard about it. Want to be my ghostwriter?

Ognyan Kulev added that mach documentation could be retried using:

wget -m -np ( (nearly 250M)
wget -m -np ( (nearly 2G (e.g. includes oskit))
wget -m -np ( (About 20-30M)

and said "Look at ( . I like Mach 3 Kernel Principles and Mach 3 Servers' Writer very much."







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 All pages on this site are copyright their original authors, and distributed under the terms of the GNU General Public License version 2.0.