...

And fails, I guess?

Yes, of course.

...

Any ideas how to solve this issue?

We need a self-contained C program that reproduces this behaviour - but I fear it's difficult or even impossible to get this...

Given that Kylix is closed source (commercial) I do not see how I could produce a self-contained C program that duplicates the behaviour. I don't know if it's an option, but there is a free 'open edition' of Kylix available for download... Can I do something else to help solving this issue? Perhaps supplying a compiled console-only program? BTW: The linux32 --3gb does not help :-( Thanks, Willibald P.S.: I can not find the ltrace32 tool mentioned in the documentation on my system or on the DVD..We need a self-contained C program that reproduces this behaviour -

Hello! Today I tried to run a 'Hello world' console-only program that I did in Kylix on SuSE 8.0(?). Unfortunately it did not run on SuSE 9.0 AMD64 because Kylix uses a function that includes following call: IconvContext := iconv_open(nl_langinfo(_NL_CTYPE_CODESET_NAME), 'UNICODELITTLE'); The problem here is, that IconvContext is always -1, which leads to a runtime error in Kylix based applications. The constants are defined as LC_CTYPE = 0; _NL_CTYPE_CODESET_NAME = LC_CTYPE shl 16 + 14; libc = 'libc.so.6'; The functions map to function nl_langinfo(item: integer): pchar; cdecl; external libc name 'nl_langinfo'; and function iconv_open(ToCode: PChar; FromCode: PChar): Integer; cdecl; external libc name 'iconv_open'; I hope the linker knows that a 32bit program needs the 32bit library, if not specified otherwise... But I guess that works, because of ldd Project2 /lib/libNoVersion.so.1 => /lib/libNoVersion.so.1 (0x40017000) libpthread.so.0 => /lib/libpthread.so.0 (0x40030000) libdl.so.2 => /lib/libdl.so.2 (0x40083000) libc.so.6 => /lib/libc.so.6 (0x40086000) /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000) Any ideas what's going on here?

Run strace on it and check what it does. Run ltrace and look at it...

"strace -cfirvx -o dccstrace ./dcc" just gives % time seconds usecs/call calls errors syscall 100.00 0.000000 0 total When starting the IDE with strace, I get the following messages (The 64-bit process is probably due to a 'source' command in the start-script.) strace -cfirvx -o delphistrace ./startdelphi [ Process PID=10678 runs in 32 bit mode. ] [ Process PID=10657 runs in 64 bit mode. ] ./startdelphi: line 26: 10678 Speicherzugriffsfehler /home/willi/kylix3/bin/ delphi $* Actually there are syscalls that returned errors in the logfile. However, I don't know if that helps - because on my 32bit SuSE Linux there are errors too, and Kylix runs fine..

...

P.S.: I can not find the ltrace32 tool mentioned in the documentation on my system or on the DVD.

It might be on the ftp server only, I'll check next week,

Thanks. BTW: DDD has problems debugging 32bit executables and there is no ddd32 available. Have a nice weekend, Willi

Willibald Krenn writes:

Hello!

Today I tried to run a 'Hello world' console-only program that I did in Kylix on SuSE 8.0(?). Unfortunately it did not run on SuSE 9.0 AMD64 because Kylix uses a function that includes following call:

IconvContext := iconv_open(nl_langinfo(_NL_CTYPE_CODESET_NAME), 'UNICODELITTLE');

Just for reference: The problem here was that /usr/lib/gconv/gconv-modules was not there, just copying the version From lib64 solved that problem (at least for me). Andreas -- Andreas Jaeger, aj@suse.de, http://www.suse.de/~aj SuSE Linux AG, Maxfeldstr. 5, 90409 Nürnberg, Germany GPG fingerprint = 93A3 365E CE47 B889 DF7F FED1 389A 563C C272 A126

Am Sonntag, 11. Januar 2004 19:51 schrieb Andi Kleen:

You don't need ddd32. You just have to tell it to use the correct child gdb for 32bit programs:

ddd --debugger gdb32 ...

I knew there had to be some command line switch.. (Used gdb32 directly in the mean time)

Next release will have that fixed because the gdb there can debug both 32bit and 64bit from a single executable.

Excellent! Willi

On Sun, 11 Jan 2004 19:36:13 +0100 Willibald Krenn wrote:

...

Here's the kernel patch for the LDT problem if you want to test yourself (against 2.6, but the code is identical in 2.4 and 2.6)

I'm still running 2.4.x - out of two reasons: Kylix does not like the new

The patch applies to 2.4.x too.

threads found in 2.6 (AFAIK this can be 'fixed' via LD_ASSUME_KERNEL) and

It should work without that on SuSE. LD_ASSUME_KERNEL is only needed when the userland support NPTL, which SuSE 9.0 doesn't.

second, I did not get my SIL 3112A (plus Seagate A-ATA HDD) running in DMA mode when using kernel 2.6.1-mm1... (Probably I did something wrong when doing the make oldconfig, but as 2.4.x works, I've no reason to switch.)

And it worked with 2.4? That's a big strange. What driver did you use in 2.4? -Andi

Willibald Krenn writes:

Only one thing gets a little bit on my 'nerves': My PCI Radeon 9200 SE is not recognized by the X-Server and the ATI driver seems to be 32bit only. :-(

configure it as 9200...

(Still hoping that SuSE will provide an update when the next version of XFree with Radeon 9200 support comes out)

I also had some issues with - I think - USB (some kernel Ooopses on system startup), but after the kernel update and disconnecting the USB plug of my MS Natural Keyboard the system is running very well [for a Linux-Box ;-)] with the Logitech mouse as sole USB device.

I've one criticism though: Unfortunately the DVD does not support a menu point saying 'Re-Install original Kernel plus bootloader'. I really would have needed this at one point, because I completely blew my grub config. (2.6.1-rc3 was not booting and somehow I had 'destroyed' the 2.4 backup) I was badly surprised when I found out that the manual repair dialog did not feature a 'Re-Install kernel plus bootloader' option and the automatic repair did not work. Another note: I had to choose manual installation, as the automatic one leads to Ooops(?).

Try "idle=poll" added to the automatic installation command line. There're buggy biosses out there...

If I had to give SuSE 9.0 AMD64 a mark (1..5 - 'excellent'...'forget it') I would probably give it a 3. (Mostly works, but nowhere as unproblematic as it should be.) However, I have to say that I'm deeply impressed by the support in this list!!

So I guess that gives a 2 at last. :-)

Thanks, Andreas -- Andreas Jaeger, aj@suse.de, http://www.suse.de/~aj SuSE Linux AG, Maxfeldstr. 5, 90409 Nürnberg, Germany GPG fingerprint = 93A3 365E CE47 B889 DF7F FED1 389A 563C C272 A126

On Sun, 11 Jan 2004 22:38:27 +0100 Willibald Krenn wrote:

...

Am Sonntag, 11. Januar 2004 19:53 schrieben Sie:

...

[Sil 3112A + S-ATA ST3120026AS in UDMA) And it worked with 2.4? That's a big strange. What driver did you use in 2.4?

Don't know what driver SuSE 9.0 uses as default, but I guess it's 'ataraid' (even if I don't have a RAID setup yet in SIL BIOS)

ataraid doesn't work on 2.6. Known problem. The driver still needs to be ported.

...

I also had some issues with - I think - USB (some kernel Ooopses on system startup), but after the kernel update and disconnecting the USB plug of my MS Natural Keyboard the system is running very well [for a Linux-Box ;-)] with the Logitech mouse as sole USB device.

It's most likely a bug in the shuttle BIOS. Try updating the BIOS. Or disable "USB legacy support" in the BIOS setup (may or may not help)

You can test that by booting the kernel with idle=poll (not recommended as a permanent solution because it burns a lot of power). If that helps the BIOS is missing a workaround for an CPU bug that causes it to corrupt 64bit registers when running the USB emulation code.

...

Only one thing gets a little bit on my 'nerves': My PCI Radeon 9200 SE is not recognized by the X-Server and the ATI driver seems to be 32bit only. :-( (Still hoping that SuSE will provide an update when the next version of XFree with Radeon 9200 support comes out)

In the worst case you can run a 32bit X server as a workaround. That should work.

...

I've one criticism though: Unfortunately the DVD does not support a menu

I do sympathise! I just purchased an AMD64fx51 system on an Asus sk8n motherboard which has a promise sata raid setup. It was purchased with XP and a striped array on sata drives already setup. It cost me close on £3000 with a 21" monitor and lots of firepower! I want Linux of whatever flavour and SuSE looked to be supporting the AMD64. I also looked at red hat but after 9 it seems that I would have to use fedora, and 64bit support seems a little way off becoming stable. Mandrake was another contender and appeared to support the promise interface. Naively I thought SATA and Raid would be working. I could see from past experiences, that at the outset there might be support issues, just as when I bought my first PII400 with an ATI all in wonder tv card - it also took some time to be properly supported. I looked at SuSE and balked at the price, but hey it might be worth it. I e-mailed SuSE uk and expressed my concerns over my 2month old system not being compatible. They kindly suggested I try the 1st cd of SUSE 9.0 and sent it to me in the post. Both this and Mandrake and RH9.0 exhibit the same problem of not recognising the devices which I presume are the drives, remembering of course that they are formatted as a sata raid array under xp. I can't complain about SuSE, as a potential first time customer they have been very helpful, but on this particular matter, until ASUS or promise or the kernel support my kit, then a true 64bit Linux os has to be put on hold. The old bit of latin someone told me one of 'caveat emptor' which I think means buyer beware holds true. Doubtless in a month or two this support will arrive. I know that this is no consolation, but we do have a great bit of kit, and a potentially great OS. Patience ? BTW It is almost certain that I will eventually end up buying SuSE 9.0 so in my case customer service and customer satisfaction go hand in hand.... ----- Original Message ----- From: "Andi Kleen" To: "Willibald Krenn" Cc: Sent: Monday, January 12, 2004 8:36 AM Subject: Re: [suse-amd64] Sil 3112A + ST3120026AS UDMA2 point

...

saying 'Re-Install original Kernel plus bootloader'. I really would have needed this at one point, because I completely blew my grub config. (2.6.1-rc3 was not booting and somehow I had 'destroyed' the 2.4 backup) I was badly surprised when I found out that the manual repair dialog did not feature a 'Re-Install kernel plus bootloader' option and the automatic repair did not work.

Well, the idea is when you do things with your grub configuration or your kernel rpm you should be also able to fix it manually from a repair system. The automatic options are more aimed at people who don't change intimate details of the system. Obviously they cannot undo all the damage an user is able to do by hand.

...

Another note: I had to choose manual installation, as the automatic one leads to Ooops(?).

In general every time you have oopses you should report them. Otherwise they cannot get fixed.

-Andi

-- Check the List-Unsubscribe header to unsubscribe For additional commands, email: suse-amd64-help@suse.com

Am Montag, 12. Januar 2004 09:36 schrieb Andi Kleen:

On Sun, 11 Jan 2004 22:38:27 +0100 [..] ataraid doesn't work on 2.6. Known problem. The driver still needs to be ported.

Thanks for pointing that out.. Well, as everything is now working with the patched SuSE 2.4 kernel (even Kylix IDE!) I've absolutely no reason to switch to 2.6.

It's most likely a bug in the shuttle BIOS. Try updating the BIOS. Or disable "USB legacy support" in the BIOS setup (may or may not help)

True, after disabling 'USB Keyboard support' (buried somewhere deep in the BIOS config) even the automatic installer managed to boot from DVD!

You can test that by booting the kernel with idle=poll (not recommended as a permanent solution because it burns a lot of power). If that helps the BIOS is missing a workaround for an CPU bug that causes it to corrupt 64bit registers when running the USB emulation code.

Hmm, I hope Shuttle will provide an update for this. But on the other hand it's not a complete disaster because my keyboard has a PS/2 and an USB plug, so I have the choice..

In the worst case you can run a 32bit X server as a workaround. That should work.

I'll keep that in mind - if I'll somewhen need hardware acceleration this will probably the only way to go.

Well, the idea is when you do things with your grub configuration or your kernel rpm you should be also able to fix it manually from a repair system.

I understand. However I thought the install DVD would also feature such a minimal repair system. In the old SuSE days (5.3 etc.) there always was the option to re-install the kernel IIRC..

The automatic options are more aimed at people who don't change intimate details of the system. Obviously they cannot undo all the damage an user is able to do by hand.

However, I'm still of the opinion that it would be very helpful if the install DVD featured a package reinstall/delete/.. dialog without having to go through all the disk-scans etc.

In general every time you have oopses you should report them. Otherwise they cannot get fixed.

Seems this oops was related to the USB bug; Nevertheless, see below: Thanks, Willi USB-Keyboard support turned ON in Shuttle BIOS (AN50S00Y): ================================================== Display 1:

...

...

SuSE Linux installation program v 1.5 (c) 1996-2003 SuSE Linux AG <<< Starting hardware detection...

Display 3: [...] insmod usb-ohci Using /modules/usb-ohci.o Display 4: [...] usb-ohci.c: usb-00:02.1, PCI device 10de:00d7 Unable to handle kernel paging request at virtual address 000000008010ddbf printing rip: 000000008010ddbf PML4 4e43067 PGD 0 Oops: 0010 CPU 0 Pid: 0, comm: swapper Not tainted RIP: 0010:[<000000008010ddbf>] RSP: 0000:ffffffff8048dfd8 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffffffff8010dda0 RCX: 0000000000000001 RDX: ffffffff80479620 RSI: 0000000000000032 RDI: 0000000000000019 RBP: 0000000000000000 R08: ffffffff80488d00 R09: 0000000000000001 R10: 0000010037b8cb80 R11: 0000000000000090 R12: 0000000000000000 R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000 FS: 0(0) GS:ffffffff80484c80(0000) knlGS:0 CS: 0010 DS: 0018 CR0: 000000008005003b CR2: 000000008010ddbf CR3: 00000000101000 CR4: 6e0 Process swapper (pid:0, stackpage=ffffffff8048d000) Stack: ffffffff8048dfd8 0 0 ffffffff8010de2a 10a ffffffff8048f775 ffffffff803e4ca0 0 Call Trace: [<ffffffff8010de2a>]{cpu_idle+42} Code: Bad RIP value <0>Kernel panic: Attempted to kill the idle task! In idle task - not syncing

Hello Willibald Krenn, you told that your kylix IDE is running. I got the Problem that my kylix IDE is running too, but I always get a error if compile a program with kylix. The error is: [Fehler] Falsches Dateiformat: libX11.so.6 I installed kylix with linux32 setup.sh and i start kylix with linux32 startdelphi my system is a suse9/amd64 kernel 2.4.21-201-default any hint for me ? thank you detlef oertel Willibald Krenn wrote:

Am Montag, 12. Januar 2004 09:36 schrieb Andi Kleen:

...

On Sun, 11 Jan 2004 22:38:27 +0100 [..] ataraid doesn't work on 2.6. Known problem. The driver still needs to be ported.

Thanks for pointing that out.. Well, as everything is now working with the patched SuSE 2.4 kernel (even Kylix IDE!) I've absolutely no reason to switch to 2.6.

...

It's most likely a bug in the shuttle BIOS. Try updating the BIOS. Or disable "USB legacy support" in the BIOS setup (may or may not help)

True, after disabling 'USB Keyboard support' (buried somewhere deep in the BIOS config) even the automatic installer managed to boot from DVD!

...

You can test that by booting the kernel with idle=poll (not recommended as a permanent solution because it burns a lot of power). If that helps the BIOS is missing a workaround for an CPU bug that causes it to corrupt 64bit registers when running the USB emulation code.

Hmm, I hope Shuttle will provide an update for this. But on the other hand it's not a complete disaster because my keyboard has a PS/2 and an USB plug, so I have the choice..

...

In the worst case you can run a 32bit X server as a workaround. That should work.

I'll keep that in mind - if I'll somewhen need hardware acceleration this will probably the only way to go.

...

Well, the idea is when you do things with your grub configuration or your kernel rpm you should be also able to fix it manually from a repair system.

I understand. However I thought the install DVD would also feature such a minimal repair system. In the old SuSE days (5.3 etc.) there always was the option to re-install the kernel IIRC..

...

The automatic options are more aimed at people who don't change intimate details of the system. Obviously they cannot undo all the damage an user is able to do by hand.

However, I'm still of the opinion that it would be very helpful if the install DVD featured a package reinstall/delete/.. dialog without having to go through all the disk-scans etc.

...

In general every time you have oopses you should report them. Otherwise they cannot get fixed.

Seems this oops was related to the USB bug; Nevertheless, see below:

Thanks, Willi

USB-Keyboard support turned ON in Shuttle BIOS (AN50S00Y): ==================================================

Display 1:

...

...

...

SuSE Linux installation program v 1.5 (c) 1996-2003 SuSE Linux AG <<<

Starting hardware detection...

Display 3: [...] insmod usb-ohci Using /modules/usb-ohci.o

Display 4: [...] usb-ohci.c: usb-00:02.1, PCI device 10de:00d7 Unable to handle kernel paging request at virtual address 000000008010ddbf printing rip: 000000008010ddbf PML4 4e43067 PGD 0 Oops: 0010 CPU 0 Pid: 0, comm: swapper Not tainted RIP: 0010:[<000000008010ddbf>] RSP: 0000:ffffffff8048dfd8 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffffffff8010dda0 RCX: 0000000000000001 RDX: ffffffff80479620 RSI: 0000000000000032 RDI: 0000000000000019 RBP: 0000000000000000 R08: ffffffff80488d00 R09: 0000000000000001 R10: 0000010037b8cb80 R11: 0000000000000090 R12: 0000000000000000 R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000 FS: 0(0) GS:ffffffff80484c80(0000) knlGS:0 CS: 0010 DS: 0018 CR0: 000000008005003b CR2: 000000008010ddbf CR3: 00000000101000 CR4: 6e0 Process swapper (pid:0, stackpage=ffffffff8048d000) Stack: ffffffff8048dfd8 0 0 ffffffff8010de2a 10a ffffffff8048f775 ffffffff803e4ca0 0 Call Trace: [<ffffffff8010de2a>]{cpu_idle+42}

Code: Bad RIP value <0>Kernel panic: Attempted to kill the idle task! In idle task - not syncing