Hurd Traffic #10 For 10 Aug 1999

By Zack Brown

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


This is the first week that a significant amount of the debian-hurd list has not been covered in KC Debian Hurd. There were a lot of posts and a good number of threads. I would have liked to cover them all, but this week I've been very busy with other things. I'll try to make up for it next week.

Mailing List Stats For This Week

We looked at 67 posts in 188K.

There were 24 different contributors. 10 posted more than once. 9 posted last week too.

The top posters of the week were:


1. Keyboard Remapping
28 Jul 1999 - 1 Aug 1999 (4 posts) Archive Link: "Ewww! Qwerty!"
People: Marcus BrinkmannKirstin S. Reese

Kirstin S. Reese asked where 'loadkeys' was, so he could remap his keyboard to Dvorak. Marcus Brinkmann replied:

There is no standard user mechanism to change the keymap.

There is a GNU Mach call to change the keymap, and there are some example programs in the mailing list archive and in my archive of the GNU hurd related lists.

I am happy to send the information to a developer who is willing to write a fancy user program similar to loadkeys or port loadkeys to the Hurd. It would be best to take a look at the linux console-* packages and to use their data files. If nobody volunteers, the information will end on the web page someday, but I have not yet extracted it from the lists (that's why I would only send it now if someone really wants to write, otherwise I will postpone this).

Any volunteers? To the Hurd developers: Should this become an external program or part of the Hurd utils?


2. Hurd Success; Reviewing Boot Messages; Cross-Install Bug Report
30 Jul 1999 (2 posts) Archive Link: "Debian Hurd installation Questions"
People: Marcus BrinkmannReinhard VicinusMatthew Vernon

Reinhard Vicinus reported ecstatic success with the Hurd, but had some problems. He had run into the 1-gig-per-partition limit, as Marcus explained (and invited Matthew Vernon to give a warning about the problem in the Easy Guide). Reinhard also asked if it was possible to review boot messages (e.g. with 'dmesg'), but Marcus explained, "not yet, unfortunately. Someone needs to do something about the kernel logger and also port the Linux syslog daemon." But he added, "If you got messages about not detected hardware that's okay. You don't have all the hardware for that support is compiled into the kernel. There are some other error messages at boot time that are not considered to bne harmful."

Reinhard also reported, "init try to run at boot time /libexec/runsystem but there is only a file /libexec/runsystem.gnu and after I copied runsystem.gnu to runsystem I got no error messages. ( Ok, it was try and error, but it worked. I hope :-/ )" Marcus replied, "Argh, okay, that's a bug in cross-install, it doesn't do the right thing here... I forgot to update it. Thanks for pointing out, will be fixed immediately."


3. Compressing gnumach And serverboot; How To Name Them
30 Jul 1999 (3 posts) Archive Link: "naming of gzip'ed gnumach and serverboot"
Topics: Bootloaders
People: Roland McGrath

Marcus announced that to save disk space, gnumach and serverboot would be gzipped in their future packages. Apparently GRUB had no problem supporting that, although Marcus added that everyone would have to upgrade their GRUB once the change took place.

He wanted some opinions on how to name the compressed files, whether to use gnumach.gz and serverboot.gz (which would be more intuitive), or keep their original names unchanged (which would not require changing the GRUB menu.lst files). Personally, he preferred the first solution, since it was a one-time inconvenience, while the second would have long-term influence. Roland McGrath and Ed Boraas agreed, and Ed added that Hurd development was still early enough that such a change would not be out of place.


4. The One-Gig Limit
30 Jul 1999 - 1 Aug 1999 (6 posts) Archive Link: "Install attempt"
Topics: FS: ext2
People: Marcus BrinkmannOlivier GalibertMichael Bacarella

Scott Jacobsen installed the Hurd for the first time. He'd used gnu-19990104.tar.gz, which seemed old to him, and Marcus said there was a gnu-19990505.tar.gz in the same directory as the old one, and which was much more current.

Scott also reported running on a 6 gig partition without any complaints, and wondered if this was too big (he'd heard about the 1 gig limit). This came as a big surprise to Marcus, who said, "MUCH to big. It did really not complain? Note that we talk about the partition, not about the hard disk as a whole. Are you sure you made a 6GB ext2fs partition for the Hurd? The partition should be ~ 1 GB max." .

Michael Bacarella was fed up with the 1 gig limit, and suggested gathering information about it, so that some effort to fix it could be launched. Marcus said that very little could be done if no developer came forward to work on it, but he agreed that information should be available; and started wading through the mailing list archives for any scraps to put up on the web page.

Olivier Galibert said, "AFAICT, the main problem is that libdiskfs mmap()s the whole partition in memory." Michael guessed, on the other hand, that it might be the limits of the data types used.


5. Longstanding Bug Tracked Down
30 Jul 1999 - 1 Aug 1999 (4 posts) Archive Link: "Few questions"
Topics: Apt
People: Marcus BrinkmannKalle Olavi Niemitalo

Michel Banguerski was seeing only half of his keys echoed to the terminal, and he had to type all his commands twice to get them to work. Marcus replied, "I see this problem only after running native-install, and when I rreebboooott :), everything is okay again."

Michel also asked about the status of the apt-get port, and Marcus explained:

Unported. I will work on it again, but it will take some time to get it done right. Also, apt can be dangerous because it may try to install packages that must be avoided (makedev, login), and I have not yet examined how confused it may get.

Mass upgrades are not good yet. It is better to install/update one package at a time. This also helps us to detect and record problems (dpkg has no useful logging capabilities I know of).

The problem with keys having to be typed twice came up again under the Subject: Could we fix rreebboooott? ( , when Kalle Olavi Niemitalo noticed that "there were two /hurd/term daemons reading the Mach console device: one for /tmp/console and the other for /dev/console."

Marcus replied, "Ah, that's the problem... seems we have found a way to provide two virtual consoles... on the same screen! ergh :)"


6. Perl; Timezones; Gratuitous Error
31 Jul 1999 - 1 Aug 1999 (2 posts) Archive Link: "hurd multiple questions..."
People: Marcus Brinkmann

Peter Hawkins reported success installing the Hurd, but had some problems. In particular, the recent gutting of the Perl packages had bitten him, but Marcus explained that this had been fixed for a few days, and the scripts had been changed to all use the current Perl version.

Peter also suggested that "timezones" be taken off the required-packages list, and Marcus agreed.

Peter was also seeing a "gratuitous error" when doing 'ls -l /dev', and Marcus explained, "That's because we can't provide the correct information for the /dev/fd translator. Instant faking wrong information, we just return a gratuitous error message. I think this ratuitous error is not correct, and the behaviour of the fd translator should be changed to provide as much useful information as possible, fake some information that doesn't harm and return a useful error message in all other cases. But I am afraid that for this special situation no error number in glibc is appropriate (do we need a new error number for this?)"


7. Architecture Independence In Debian
3 Aug 1999 - 5 Aug 1999 (3 posts) Archive Link: "Q: binary-all group"
People: Kalle Olavi NiemitaloMarcus Brinkmann

Peter Cejchan asked if main/binary-all contained architecture-independent files, and if so, what made them architecture independent. Kalle Olavi Niemitalo replied:

File formats. Binary-all packages contain files in architecture-independent formats like Perl, PostScript, HTML, manual pages, plain ASCII text, GNU message catalogs or PNG. Executable files are usually architecture-dependent because they contain machine instructions which only one architecture can easily execute, but scripts are architecture-independent.

When a binary-all package is built from a source package, the files are often copied without any compiling. But there are exceptions: Texinfo is compiled to Info, PO is compiled to GMO and so on. These formats have been defined such that all kind of machines can use them.

The GNU Coding Standards recommend making all data files architecture-independent.

And Marcus added:

The current situation is that binary-all means: Works on all Linux ports (i386, m68k, etc). What I would like to mean it is: Works on all Debian machines, Linux and the Hurd. Hence, we need a new section binary-linux-all and binary-hurd-all. Ideally, we would have much more sections (for example binary-all-i386). Unfortunately, only a few packages are affected, and the people who could implement the changes in the archive etc are very busy and would need to be convinced that the clean solution is better than making for example makedev architecture dependant.

If you don't understand a word, that's okay, because it is not your problem, it is ours. For our users is only relevant that

  1. binary-all is filled with linux software for historical reasons and
  2. some of this linux software is downright HARMFUL under the Hurd, like the makedev package.

For now, we will add Conflicts: for harmful packages so they are not installed by accident.







We Hope You Enjoy Hurd Traffic

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.