Hurd Traffic #4 For 29 Jun 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

• Printer-Friendly Format
• Text Format
• XML Source
• Mailing List Stats For This Week
• Threads Covered

  1.	13 Jun 1999 - 14 Jun 1999	(3 posts)	Grub Extended Partition Limitations
  2.	15 Jun 1999 - 21 Jun 1999	(5 posts)	GMP Renovations
  3.	15 Jun 1999 - 17 Jun 1999	(3 posts)	Ongoing 'ltrace' And 'ptrace' Efforts
  4.	17 Jun 1999 - 20 Jun 1999	(5 posts)	More Grub Trouble And Success
  5.	17 Jun 1999 - 24 Jun 1999	(5 posts)	System Clock Limitations
  6.	17 Jun 1999 - 20 Jun 1999	(2 posts)	Partition Sizes
  7.	17 Jun 1999 - 18 Jun 1999	(2 posts)	'/var/run/mtab', Floppy Mounting, And 'minicom' Questions
  8.	18 Jun 1999	(1 post)	Hurd 19990616 And Gnumach 1.1.92-5 Released
  9.	19 Jun 1999 - 22 Jun 1999	(16 posts)	More On NFS And Translators
  10.	20 Jun 1999	(1 post)	More On Profiling
  11.	20 Jun 1999 - 21 Jun 1999	(10 posts)	More On Lists Of Easy Packages
  12.	20 Jun 1999 - 21 Jun 1999	(4 posts)	Maintaining An Existing Hurd System
  13.	21 Jun 1999 - 24 Jun 1999	(3 posts)	PCMCIA Itchy Fingers
  14.	21 Jun 1999 - 22 Jun 1999	(7 posts)	Corrupt Files Tracked Down

Mailing List Stats For This Week

We looked at 81 posts in 242K.

There were 29 different contributors. 16 posted more than once. 12 posted last week too.

The top posters of the week were:

• 12 posts in 35K by Marcus Brinkmann <Marcus.Brinkmann@ruhr-uni-bochum.de>
• 12 posts in 30K by tb@MIT.EDU (Thomas Bushnell, BSG)
• 5 posts in 15K by Roland McGrath <frob@debian.org>
• 4 posts in 17K by Allover Stripes <allover@hedgehog.dyn.ez-ip.net>
• 4 posts in 15K by Momchil 'Velco' Velikov <velco@fadata.bg>
• Full Stats

1. Grub Extended Partition Limitations

13 Jun 1999 - 14 Jun 1999 (3 posts) Subject: "GRUB and logical partitions? (ver 0.5.91)"

Topics: Bootloaders

People: Alexander K. Hudek, OKUJI Yoshinori

Alexander K. Hudek had some problems with grub recognizing his logical partitions. OKUJI Yoshinori said there shouldn't be a problem with it, and under the Subject: GRUB and logical partitions followup, Alexander reported, "GRUB only checks for extended partition type 0x5. My system had an extended partition of type 0xf. Debian fdisk describes this as a Win95 (LBA) extended partition. I retyped it to five and it work fine (Course maybe my other software doesn't ...like partition magic in win98). At any rate GRUB should probably be patched to recognize type 0xf as extended as well (to whoever maintains GRUB now)."

2. GMP Renovations

15 Jun 1999 - 21 Jun 1999 (5 posts) Subject: "gmp2"

People: Niels Möller, Pavel Roskin, Marcus Brinkmann

Per Lundberg reported that the gmp2 (arbitrary precision arithmetic library) package seemed broken, in particular it seemed to be linked to Linux libc6. Marcus Brinkmann confirmed the breakage, and suggested Per compile and upload a fixed version. Per tried, and while he didn't notice any problems compiling, the resulting libraries choked on some test data.

Pavel Roskin replied that the Makefiles were hardcoded for Linux; he added that the whole package should be seriously modified to use 'autoconf', 'automake' and 'libtool', and said he'd give it a try.

Niels Möller gave a link to the gmp homepage, and suggested talking to Torbjörn Granlund about migrating to autoconf et al. He added that autoconfing had been on the TODO list for awhile, and also explained, "Note that autoconfing gmp is probably a little different from most other packages, ideally you will want to do some performace tests to aid choice of algorithms etc. In my opinion, when changing the build stuff, autoconfiscating should have a _lot_ higher priority than converting the makefiles to use automake and libtool." He added that there was an autoconfed version of gmp in the ssh-1.2.x packages, but he hadn't checked it out.

3. Ongoing 'ltrace' And 'ptrace' Efforts

15 Jun 1999 - 17 Jun 1999 (3 posts) Subject: "More ptrace/hurd questions"

People: Roland McGrath, Brent Fulgham

Continuing the saga from last week in Issue #3, Section #8  (8 Jun 1999: Porting Debugging Tools To The Hurd) , Brent Fulgham had been delving into the ltrace and hurd internals, to try to get ltrace and ptrace working. ltrace actually compiled, but would die immediately with a memory access error.

He had hacked around with gdb, but was still confused about what he was looking for. Roland McGrath replied that Brent was being much too vague, and should post more specific details. Under the Subject: Additional ltrace detail, Brent posted the output of a big 'gdb' session, but there was no reply.

4. More Grub Trouble And Success

17 Jun 1999 - 20 Jun 1999 (5 posts) Subject: "Installation problems with GRUB"

Topics: Bootloaders

People: Kalle Olavi Niemitalo

Shane McDonald was trying to set up a triple-boot system with Windows 95, Debian Linux (slink), and the Hurd. The installation went fine, but booting with 'grub' gave the error "Filesystem type unknown, partition type 0x55".

Kalle Olavi Niemitalo replied:

Shane gave it a try and reported complete success: He rebooted with the Windows 95 boot disk, used DOS fdisk to remove the partition on the second drive, added a new partition, and reinstalled the Hurd, after which 'grub' worked. End Of Thread.

5. System Clock Limitations

17 Jun 1999 - 24 Jun 1999 (5 posts) Subject: "Some issues."

People: Roland McGrath

Fabricio Chalub reported a successful Hurd installation. He asked a couple questions, one of which was that timezone information was either broken or handled differently from Linux. There was no reply to that post, but 5 days later under the Subject: Time problems., he gave a more complete summary of the problem.

Linux and the Hurd both had /etc/localtime pointing to /share/zoneinfo/Americas/Sao_Paulo, and /etc/timezone had the same string "Americas/Sao_Paulo", under both Linux and the Hurd. However, he was noticing that files created under Linux, became 3 hours older when he booted into Linux.

He added that his system clock was set to local time, and his guess was that Linux understood that, while the Hurd assumed GMT time. Roland McGrath concurred, and said, "Machine clocks ought to be in universal time, and the Hurd doesn't actually support anything else at the moment. The only reason to keep your machine clock in local time is because DOS and Windows like it that way, but you wouldn't run one of those systems, now would you?" and added, "Seriously, we will probably add a way to tell the system that your machine clock is in local time, but it is not clear off hand how to do it and it will definitely be a low priority."

6. Partition Sizes

17 Jun 1999 - 20 Jun 1999 (2 posts) Subject: "Re: my experiences"

Topics: FS: ext2

People: OKUJI Yoshinori, Marcus Brinkmann

Continuing the saga from Issue #3, Section #14  (9 Jun 1999: Installation Difficulties) , Tim Newsham (who started the original thread), said that the error he'd been seeing,

hd02: badd access: block=28 count=2
end_request: I/O error, dev 03:02 sector 28
Hurd Server bootstrap: ext2fs.static[hd0s7] exec

was only due to his Hurd partition being >1G. He resized it and had a successful install, with an apparently harmless error message at boot (which he did not specify). OKUJI Yoshinori, who had also been working on that puzzle, asked if Tim had tried his patch, or the new gnumach package from Marcus Brinkmann; and if so, asked if Tim would post the boot error he was seeing.

There was no reply this week.

7. '/var/run/mtab', Floppy Mounting, And 'minicom' Questions

17 Jun 1999 - 18 Jun 1999 (2 posts) Subject: "minicom & mtab questions"

Topics: Bootloaders, FS: ext2

People: Roland McGrath

Andrew Comech asked how to handle /var/run/mtab: if the file didn't exist, e2fsck gave errors, while if it did exist, it would go away after cleaning /var/run. Roland McGrath explained that the Hurd didn't use an mtab file; it was there only to satisfy limitations in 'e2fsck'. He didn't know why mtab was getting cleaned out of /var/run.

Andrew also had trouble mounting floppies. 'settrans /floppy /grub/ext2.fs /dev/fd0' seemed to work at first; but when he tried to actually access the device, the system would freeze until he hit ^C. Roland recommended, "Use `showtrans /floppy' to see what the current passive translator setting is. To try setting the active translator interactively so you can see any error messages it might have, do `settrans -agf /floppy /hurd/ext2fs'. If you really have been using `/grub/ext2.fs' for some reason (I don't know what you think this has to do with GRUB), and that is not a place you have installed the program usually installed in /hurd/ext2fs, then that could be your problem."

Finally, Andrew also had minicom problems. Apparently, 'minicom -s' gave the usual configuration menu, but when exiting, it would say "minicom: stopped" just as if no '-s' had been given. Without strace, he didn't know what to do, and Roland suggested he run minicom under 'gdb'.

8. Hurd 19990616 And Gnumach 1.1.92-5 Released

18 Jun 1999 (1 post) Subject: "update of hurd and gnumach"

People: Marcus Brinkmann

Marcus Brinkmann announced:

There were no replies.

9. More On NFS And Translators

19 Jun 1999 - 22 Jun 1999 (16 posts) Subject: "Re: NFS -- Hurd"

Topics: FS: NFS

People: Kalle Olavi Niemitalo, Thomas Bushnell, Gordon Matzigkeit, Roland McGrath

Continuing the saga from Issue #2, Section #7  (2 Jun 1999: The Hurd Over NFS) , in response to Roland McGrath's statement that it was not currently possible to have a Hurd system entirely diskless, Kalle Olavi Niemitalo asked, "When it is, where will the NFS client save passive translators?"

Thomas Bushnell replied, 16 days later, "My best answer is that NFS will only support passive translators by extending the protocol, and then it will only do anything if the server supports it. There are gross hacks that could make a Hurd-savvy client deal with a plain vanilla nfs server, but that's not in the cards just yet."

There followed an interesting discussion of the problems of NFS support of passive translators; Wish I Understood More (tm). Apparently the solutions involve various complex indirections, but most folks had at least some idea that it would ultimately be possible, though ugly.

Interesting tidbit: Andrew Archibald asked how hard it was to write a translator. He recalled that Gordon Matzigkeit had written a translator development tool a few months ago, and wondered if it was available. On the difficulty question, Gordon replied:

On the development tool, he gave a pointer to ftp://alpha.gnu.org/gnu/trans-0.0a.tar.gz and added that it would soon be resurrected under the name 'metafs'.

10. More On Profiling

20 Jun 1999 (1 post) Subject: "Re: Hurd Latencies -- Questions..."

People: Thomas Bushnell

Continuing the thread from last week in Issue #3, Section #10  (8 Jun 1999: Hurd Profiling) , Thomas Bushnell added, "no serious time has been spent in profiling. It would be helpful to profile some of the more important servers, the C library, and the kernel. Doing the C library and the kernel is a little trickier than just doing gprof, but the servers should be gprofable without much effort. Then we can know where to look."

11. More On Lists Of Easy Packages

20 Jun 1999 - 21 Jun 1999 (10 posts) Subject: "Re: NCFTP package"

People: Pavel Roskin, Thomas Bushnell, Roland McGrath, Marcus Brinkmann

Continuing the thread from Issue #3, Section #6  (8 Jun 1999: Successful Hurd Install; Installing A Bootloader; Mounting A CD; Removing A Translator; ncftp Success; Lists Of Good Packages; apt-get Saga) , where Marcus Brinkmann had posted several incomplete lists of packages that compiled with little or no effort; this week Thomas Bushnell pointed out that the list at http://master.debian.org/~brinkmd/hurd/package-list.html was well out of date. Marcus acknowledged the problem, and promised to fix it, and make the new list available at http://www.debian.org/ports/hurd

Pavel Roskin pointed out that even the new lists contained a number of packages for which more recent releases were available (one exception being mawk-1.3.3-2 -- Red Hat shipped mawk-1.2.2). Marcus replied that there were internet connection charges to consider, and that anyway the problem was irrelevant because the porting problems were present in the latest versions of the packages as well.

In the course of discussion, it came out that the Hurd's svgalib package was a dummy. Roland McGrath felt it should be possible to port svgalib without too much trouble, and Marcus invited someone to do it.

12. Maintaining An Existing Hurd System

20 Jun 1999 - 21 Jun 1999 (4 posts) Subject: "after getting started..."

People: Roland McGrath, Marcus Brinkmann

Tim Newsham successfully installed the Hurd. He had a working system, and wanted to know where to go from there, so he posted some questions. He asked how to install more .deb files, or how to upgrade packages that were already installed. Roland McGrath said it was just like Debian: use 'dselect' and 'dpkg'. Tim also wanted to know if :pserver:anoncvs@anoncvs.gnu.org/gd/gnu/anoncvsroot was the canonical CVS repository, and Roland replied that yes, it was a read-only mirror of it, updated once per day. Tim also asked what sources he needed if he wanted to rebuild the entire system. Roland replied, "If you want to rebuild from scratch, you need the usual tools plus mig. Build glibc, then gnumach, then hurd."

Tim also asked if there was a list of who was doing what, and Marcus Brinkmann replied, "I maintain such a list for myself, but it does only cover some area. Best is to follow the debian-hurd mailing list closely. and read the archive. Urgent things and plans are always posted here."

13. PCMCIA Itchy Fingers

21 Jun 1999 - 24 Jun 1999 (3 posts) Subject: "PCMCIA support may arrive"

People: Thomas Bushnell

Thomas Bushnell said, "So I'm buying myself a spiffy laptop. That happens to mean that PCMCIA support may suddenly become a priority for me. Anyone who has any suggestions or advice, please let me know."

14. Corrupt Files Tracked Down

21 Jun 1999 - 22 Jun 1999 (7 posts) Subject: "Glibc build problems"

People: Roland McGrath

Momchil 'Velco' Velikov had a 7-post staircase with Roland McGrath over the course of about an hour and a half. Momchil was trying to compile glibc-2.1.1 on a Linux box for the Hurd, using egcs-1.1.2, gnumach-19981118, mig-1.0.2, hurd-19990329, and glibc-2.1.1, and was getting the following errors:

hurd.h: In function '__hurd_fail':
In file included from hurdstartup.c:24:
hurd.h:51: `EMACH_SEND_INVALID_DEST' undeclared (first use in this function)
hurd.h:51: (Each undeclared identifier is reported only once
hurd.h:51: for each function it appears in.)
hurd.h:52: `EMIG_SERVER_DIED' undeclared (first use in this function)
hurd.h:54: warning: unreachable code at beginning of switch statement

Roland said Momchil would have to give more information if he wanted any real help -- at the bare minimum the complete 'make' output. But he suggested that "it seems most likely that your cross-tools are not set up quite right. Take a failing gcc command and run it with -v -save-temps; the output and the # lines in the foobar.i file ought to show you where the compiler is finding all the files, and if it's getting the wrong errno.h or somesuch."

Momchil replied that the 'make' output was around 600K, but he did include the output of 'gcc -v -save-temps ...' He also noticed with some confusion that 'find . -type f -exec grep EMIG_SERVER_DIED {} \;', run from the top source dirs of glibc, gnumach, hurd and mig, didn't locate any definitions of EMIG_SERVER_DIED.

About the 'gcc' command, Roland said to look at the .i file that was produced, to see if the filenames in the # lines looked reasonable. For the failed 'find' command, he replied, "That is very odd, since it appears in sysdeps/mach/hurd/bits/errno.h in the glibc sources. Is that file missing from the distribution file?"

Momchil didn't find anything wrong in the .i file. He did have sysdeps/mach/hurd/bits/errno.h, but no EMIG_SERVER_DIED or EMACH_SEND_INVALID_DEST definitions.

Roland didn't know what to make of this. He looked around in some old and new glibc distributions, but always found those definitions where he expected them. He said:

Momchil confirmed that the problem seemed to be only on his system, and got a new glibc. End Of Thread.

Sharon And Joy