Hurd Traffic #85 For 20 Mar 2001

Editor: Zack Brown

By Paul Emsley  and  Zack Brown

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 149 posts in 944K.

There were 35 different contributors. 22 posted more than once. 12 posted last week too.

The top posters of the week were:


1. Compiling Issues, NMUs
25 Feb 2001 - 11 Mar 2001 (25 posts) Archive Link: "compiling packages"
Summary By Paul Emsley
People: Brian MayMarcus BrinkmannRobert BihlmeyerRoland McGrathIgor KhavkineJeff Bailey

When compiling cyrus-sasl Brian May encountered the following problem:

common.c:849: conflicting types for `_sasl_log'
saslint.h:184: previous declaration of `_sasl_log'

The problem was pinned down to "errno" being a macro (and thus cannot be used as a normal variable).

However, Igor Khavkine said that he had already compiled cyrus-sasl and openldap. Marcus Brinkmann told Brian and/or Igor to revise Igor's patch and report it upstream. Later Brian asked "What package do I need for threads support?" , to which Marcus answered "we only have cthreads currently" and confirmed autoconf brokenness: "It must be some version or macro problem." .

Brian got on with his compiling but found a problem at the link stage: "undefined symbol "ldbm_back_abandon"" . Marcus was quickly back with a spot-on diagnosis: "When it crashed, it left a zero length *.o file which is not rebuild (because it looks uptodate for make)" . Brian deleted the zero length object files and continued with the build, finally reporting "Success!"

Jeff Bailey asked Brian if he had posted the packages somewhere so that he could use them to build sendmail. Brian replied that he had not yet since he was "not happy yet with the results. sasl had broken Depends, and the openldap libraries were broken by libtool."

In a different subthread, there was a discussion on how to deal with binary NMUs (Non-Maintainer Uploads) in the Debian system. Robert Bihlmeyer pointed out that "the Debian Developer's Reference section 8.2 says it's okay to do binary NMUs right away. [...] That's what autobuilders are doing." Marcus cleared up a few mis-understandings, particularly on the amount of testing necessary and said "You can help everyone by submitting the bug you discover when testing it, and avoid uploading such broken packages" , and amusingly added: "[But] don't think [your] scheme works for everyone. It doesn't for me and I actually doubt it will work for you either except in the simplest of the cases. (For example, try to do a NMU of X *and* survive Brandens reactions"

Robert didn't disagree much, but said "I believe, though, that about 4500 simple cases exist (X is of course not one of these)."

Marcus finished up with: "We don't really disagree, we just imagined different cases. I should write up some check ist for common build problems which are not easy to spot if you don't know where to look for" .

In the original post Brian also asked how to deal with the MAXHOSTNAMELEN problem. Roland answered this with: "You should use the procedure specified in the libc manual for reallocating the buffer as necessary when gethostname returns ENAMETOOLONG. When there a MAXHOSTNAMELEN definition that corresponds to a fixed limit (as on most systems), then gethostname will never fail with ENAMETOOLONG. When there is no MAXHOSTNAMELEN (as on the Hurd), there is no fixed limit and theoretically it might be any size; you can pick some reasonable value like 64 or 128 for the initial buffer size, and in practice noone will use a hostname longer than that."


2. 'parted' Ported To The Hurd
6 Mar 2001 - 15 Mar 2001 (9 posts) Archive Link: "[PATCH] Operating system independence; Hurd Port."
Summary By Zack Brown
People: Andrew ClausenNeal H Walfield

Neal H Walfield posted a big patch to 'parted', and explained that by splitting OS-dependent code into their own system-specific directories, he had made parted much more portable. As a proof of concept, he'd ported it to the hurd. Andrew Clausen was very impressed, but replied:

This is a very big patch/change. I've been "playing" with 1.5.x also (after 1.5.1-pre1), mainly with file system stuff. So, I think we should fork for the time being. When the file system stuff is ready, I'll merge that in to the main branch. Fortunately, there should be no conflicts, since my work is on a different layer :-)

In future, I think it would be better to let me / the maintainer know when you intend to write a non-trivial patch. I would have preferred that this be done in lots of small patches, but don't worry now...

BTW, everything ok with assigning the copyright to the FSF?

Neal agreed with the idea of sending lots of small patches, but pointed out that he only had net access once a week, which made that a bit more difficult. For the copyright issue, Neal replied, "Yes, I have already assigned my work on the Hurd to the FSF; I assume that that should cover it. Let me know if I need to do more." To this, Andrew replied:

Your name does not appear in

So, the assignment process hasn't been completed for your Hurd work? What was the disclaimer you signed? Could you please send it to me (with what you filled in, etc.)


3. g++, libgtk1.2 And esound
8 Mar 2001 - 12 Mar 2001 (15 posts) Archive Link: "g++ on the hurd?"
Summary By Paul Emsley
Topics: FS: ext2
People: Daniel BurrowsMarcus BrinkmannRoland McGrathBen Collins

When he tried to port aptitude, Daniel Burrows found that he "had removed g++ while bringing everything up-to-date a few days ago." Marcus explained: "Something removed libstdc++2.10 (and 2.9) from the archive, and now dinstall thinks it is a new package and refuses to cooperate. [..] You can fetch the missing packages from " . Daniel tried to get the missing packages and reported that "it seems that libstdc++2.10-dev has a versioned dependency on libc6-dev. " He also listed some other breakages.

With regard to the libc6-dev issue Marcus said: " Yes. The idea was that we provide libc6-dev, but that doesn't work with versioned stuff. Please file the bug."

Daniel came back with "The logs for bug #87131 ( claim it's been fixed. I see that version of libstdc++ is the one I grabbed from incoming, so maybe it wasn't fixed after all" .

Ben Collins suggested that it would be better for libgtk1.2 not to depend on gconv-modules and Daniel filed a bug ( to this effect.

Daniel had also mentioned that he "occasionally get lots of scary-looking error messages from ext2fs about (iirc) "free inode had non-zero size.." -- could it be trying to reallocate inodes? That sounds bad.." . Marcus responded to this with "A bug in linux, if you ask me. It should clear the inode size info on filesystems owned by the Hurd. You can ignore these. If you don't ever delete files in linux from Hurd owned partitions, you won't see it."

However, Roland McGrath offered his consider opionion: "Since e2fsck doesn't fix this even for creator_os==hurd, perhaps we should avoid the complaining. But I have seen this a lot lately and I don't think I ever removed files from linux, so maybe there is another bug. "

There followed some discussion about GNOME and esound (it is a good idea for developers to have the GNOME libraries even though GNOME does not run well on the Hurd). Marcus said that esound could be compiled without sound support and Daniel replied that he managed to do that.


4. Security Issues
9 Mar 2001 - 11 Mar 2001 (8 posts) Archive Link: "Scary syslog entries, random breakage"
Summary By Paul Emsley
Topics: Apt
People: Daniel BurrowsMarcus BrinkmannOystein ViggenFrederico Muñoz

Daniel Burrows started with a report that he had compiled aptitude and libgmp3 but " the system bought the farm while I was compiling rep for the nth time. So I booted into Linux, let fsck run, checked my mail, and restarted the Hurd. The first thing I noticed was that it failed to boot; it looked almost as though it was trying to fsck and failing. The next thing I noticed was that after I booted all the way, I was unable to use the network (I got various messages, ranging from "something wicked happened" (apt) to "translator died" (ping, I believe)) This seemed a little odd, since the networking was working when I shut the system down, so I looked in the syslog for clues about why pfinet was unhappy, and [saw bizarre network connection messages]. I'm not an expert in security, but that looks like someone was trying to do something bad (the address in question is some random overseas dialup) ."

Marcus Brinkmann and Frederico Muñoz tried to reassure Daniel saying that they also had seen similar messages (and Marcus was not even connected). Frederico thought that there was a problem with the Hurd netkit and Marcus said "pfinet is sort of unstable right now" .

Later Marcus remarked: "Note that an IP stack diagnosis will show the Hurd system as a Linux 2.2 system." , which prompted Oystein Viggen to say "nmap says "No exact OS matches for host". I have now submitted the fingerpring to the nmap maintainer" .

Oystein reported that Nmap (2.54BETA7) says that the TCP Sequence Prediction was "Class=trivial time dependency, Difficulty=4 (Trivial joke)", which it was thought would make it easy to spoof against a Hurd box.


5. Status Of Apache Under The Hurd
10 Mar 2001 - 11 Mar 2001 (10 posts) Archive Link: "Reports & Questions"
Summary By Zack Brown
People: Bob HamMarcus Brinkmann

Bob Ham tried the Hurd and was very impressed by how far it had progressed since the last time he'd tried it (he didn't mention how long that had been). But he also reported, "While installing apache, it all, kind of, went a bit wrong." Marcus Brinkmann replied, "You mean this two year old broken apache package in the ftp archive? I don't think it ever worked. Someone should port the current apache Debian package to the Hurd properly, and submit the necessary changes to the BTS."


6. sshd Fun
11 Mar 2001 (5 posts) Archive Link: "sshd fun"
Summary By Paul Emsley
People: Oystein ViggenNiels MöllerRoland McGrath

Oystein Viggen has been trying to get openssh working on the Hurd. He reported "the client works flawlessly, but the server is a bit more strange" . He provided part of the server log.

Oystein asked Niels Möller for recommendations for lsh ( on the Hurd. Neils replied that he had not tried. "The only potential problem I'm aware of is the randomness initialization; recent versions of lsh try to cope without /dev/urandom by executing some random commands (vmstat, netstat, ps, etc with various flags), and seed a prng (currently based on arcfour) from the output." There followed some discussion on entropy, randomness and PRNGD ( .

Roland McGrath diagnosed the sshd output as representing a bug in pfinet. In response to the comment that there was no problem if a command was called on the server (eg: ssh otherhost cat /etc/motd), Roland said "This is not surprising. We don't know why the pfinet bug arises in some uses of TCP and not in others. Different uses of a protocol like ssh will send different amounts of data in different-sized chunks, and it is probably some quirk of just what is going through that triggers the pfinet bug. "


7. Memory Leaks
11 Mar 2001 (3 posts) Archive Link: "memory leak?"
Summary By Zack Brown
Topics: FS: ext2
People: Marcus BrinkmannDaniel Burrows

Daniel Burrows noticed that available memory gradually shrank during a Hurd session; and suspected that it might be a memory leak. Marcus Brinkmann replied, "There is at least one leak in ext2fs (or libdiskfs) I am currently investigating, there are certainly more, and probably quite a lot."


8. ftpfs
11 Mar 2001 - 12 Mar 2001 (6 posts) Archive Link: "Old message about ftpfs that works :)"
Summary By Paul Emsley
Topics: FS: FTPFS
People: Frederico MuñozThomas BushnellMarcus BrinkmannRoland McGrathOystein Viggen

Frederico Muñoz said "I was trying to get ftpfs working without much luck, and I stumbled upon an old message that contained a line that worked 100%!" . He said that the ftpfs documentation was not clear on the hostmux usage and quoted Thomas Bushnell ( who had said:

settrans -c /ftp /hurd/hostmux /hurd/ftpfs /

and then [simply use like:]

cd /ftp/

Note that Marcus Brinkmann also mentioned how to use hostmux in this post ( in this recent thread

Oystein Viggen then said that he was seeing "Gratuitous error"s using ftpfs. Marcus replied that the ftp client in ftpfs "might require some strict standard conformation, or have some bugs. You can run the translator in debug mode (see the --help output) and check the transmission log." Roland McGrath added that "The ftpfs filesystem server uses the libftpconn library to do the FTP protocol. The ftpcp and ftpdir commands (source in hurd/utils/) use this library too."

Frederico later added that he had tried to use a server running the wu ftp server, which had given him a "Gratuitous error" too. He asked how to remove the directories created by ftpfs. Oystein replied:

kill -9 all the ftpfs translators,
kill -9 the hostmuxer,
settrans -g /ftp,

but added that that was probably not the best way of doing it.


9. Compiling apt-0.5.3
12 Mar 2001 - 13 Mar 2001 (4 posts) Archive Link: "Compiling apt-0.5.3 on the Hurd"
Summary By Paul Emsley
Topics: Apt
People: Daniel BurrowsMarcus BrinkmannJeff Bailey

Daniel Burrows reported "I was trying to compile the latest apt and ran into some serious problems. It looks like apt is trying to build a library called libstdc++, and failing to build apt-pkg (I think maybe somehow apt-pkg is being named incorrectly) Unfortunately, the Makefiles are incredibly illegible and I can't figure out where this could be coming from."

"What, you don't like Jason's home brewed make system?" asked Marcus comically.

Jeff Bailey replied that in his attempts, "there appear to be 2 problems, the first problem appeared with 0.5.2, and the second one appears in 0.5.3. I haven't investigated them yet."

There appears not solution yet.


10. X Mouse Protocol, screen
13 Mar 2001 - 14 Mar 2001 (4 posts) Archive Link: "Some questions on mouse and screen"
Summary By Paul Emsley
People: Frederico MuñozMarcus BrinkmannDaniel BurrowsPaul Emsley

Frederico Muñoz kicked off the thread describing his problems of using his mice under X saying "Protocol mouseman isn't implemented AFAIK and I've lost the URL of the japanese page where the protocols were discribed. Can anybody with either a PS/2 mouse (to give me the PS/2 semantics so I can use my second mouse) or a serial logitech mouse (to give me the magic translator) help me?" . He later added "I have also checked, [...] I've found it on Kernel Cousin, thnks :)"

(ed. [Paul Emsley] You're welcome :))

Marcus Brinkmann answered this with:

You can use the native X mouse driver. Just specify

Section "Pointer"
       Protocol        "MouseMan"
       Device          "/dev/com0/index.html"

This seems to be slower though, and laggish.

Frederico said that he would try this.

Frederico also asked how to get colors when using screen. Daniel Burrows answered this, telling him to use TERM=mach-color.







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.