Skip to contentSkip to conversation

hppa binutils/glibc updates

parisc-linux Randolph Chung 10 participants 27 messages Nov 4, 2001 – Nov 7, 2001

lore
Share & tools
lore
#0 lore raw

Some good news, I hope....

With latest patches from Alan Modra, things seem to be looking better. Here's what I've tried:

In a chroot built with debootstrap plus build-essential:
- Rebuild binutils from sourceware cvs
- Using new binutils, rebuild glibc (2.2.4-5) with am's latest patch

With these bits installed, the following were tested:
- build and install wget (just a simple sanity test)
- install and run libgtk1.2 and gkrellm
- recompile and run python1.5 and apache

all these seem to be ok now. Before recompiling, things like python will die with a SIGILL

I've made a list of all the packages that got uploaded between around Oct 15 till now. All these need to be recompiled and binary-NMU's need to be made. James/LaMont/Bdale/myself/etc can do these fairly easily with sbuild after new binutils and glibc bits are uploaded. Hopefully we can also restart the buildd after that.

Chris/Ben: can you let us know when we can expect new binutils/glibc? I'll forward the glibc patch to Ben separately.

If you can think of anything else that we should try to make sure everything works, please let us know.

fingers crossed,
randolph

Signature · 4 lines
   @..@                                         http://www.TauSq.org/
  (----)
 ( >__< )
 ^^ ~~ ^^
#1 lore raw

Hello,

Has gcc been fixed to produce "working" binaries for the 720 (and perhaps other machines), with the aligned stuff being fixed.

Thanks for your hard work! :-)

Regards,

Albert

#2 lore raw
Quoted 4 lines

Has gcc been fixed to produce "working" binaries for the 720 (and perhaps other machines), with the aligned stuff being fixed.

Thanks for your hard work! :-)

Looks like the patch is in, but the new gcc hasn't yet been built because of the toolchain issues. Once the official binutils is in the archive we can build the new gcc and see if that works....

randolph

Signature · 4 lines
   @..@                                         http://www.TauSq.org/
  (----)
 ( >__< )
 ^^ ~~ ^^
#3 lore raw

Hello,

Quoted 8 lines · Randolph Chung
Quoted 4 lines

Has gcc been fixed to produce "working" binaries for the 720 (and perhaps other machines), with the aligned stuff being fixed.

Thanks for your hard work! :-)

Looks like the patch is in, but the new gcc hasn't yet been built because of the toolchain issues. Once the official binutils is in the archive we can build the new gcc and see if that works....

That's great to hear! Will someone be creating new xc tarball as well. or does this happen automagically? I need it to compile some test binaries, as I assume it's going to take many weeks for all the Debian packages to get recompiled? Or will this be unnessary in most cases?

Regards,

Albert

#4 lore raw
Quoted 4 lines · Albert Strasheim

That's great to hear! Will someone be creating new xc tarball as well. or does this happen automagically? I need it to compile some test binaries, as I assume it's going to take many weeks for all the Debian packages to get recompiled? Or will this be unnessary in most cases?

i'm pretty sure the only thing which needs to get recompiled will be glibc.

Signature · 1 line
Revolutions do not require corporate support.
#5 lore raw
In reply to Randolph Chung
Quoted 11 lines · Randolph Chung

In a chroot built with debootstrap plus build-essential:
- Rebuild binutils from sourceware cvs
- Using new binutils, rebuild glibc (2.2.4-5) with am's latest patch

With these bits installed, the following were tested:
- build and install wget (just a simple sanity test)
- install and run libgtk1.2 and gkrellm
- recompile and run python1.5 and apache

all these seem to be ok now. Before recompiling, things like python will die with a SIGILL

That's good news :-) FYI, I work at Agilent now (formerly HP-owned company) and am surrounded by PA-RISC machines. I'm trying to weasel an extra machine out of them, but have only been there two weeks so it may take some extra time :-P

Quoted 2 lines

Chris/Ben: can you let us know when we can expect new binutils/glibc? I'll forward the glibc patch to Ben separately.

I'm hoping for a new one soon. I may not package it, however, due to some flux in other platforms. If we can work together to isolate the patches that are needed for hppa, I'd be happy to include those in my next upload, though (I already am planning one).

Quoted 2 lines

If you can think of anything else that we should try to make sure everything works, please let us know.

If I think of something, I'll write ya. Also, if I manage to get a machine, I'm sure you'll be hearing from me (probably will be an Apollo series 735, fyi...slow, but effective).

C

#6 lore raw
Quoted 3 lines · Christopher C. Chimelis

If I think of something, I'll write ya. Also, if I manage to get a machine, I'm sure you'll be hearing from me (probably will be an Apollo series 735, fyi...slow, but effective).

735 is a pretty cool machine. don't get one which requires a fast-wide-diff scsi interface though; we don't have that working yet.

Signature · 1 line
Revolutions do not require corporate support.
#7 lore raw
Quoted 2 lines · Matthew Wilcox

735 is a pretty cool machine. don't get one which requires a fast-wide-diff scsi interface though; we don't have that working yet.

I think that's all we have, but I'm not sure. I'll look at the one that I do have and see what the "norm" is around the office (the one I have has to remain HP-UX for work reasons, but it should be a representative sample from what I've seen so far). These things also don't have much RAM (mine has 64MB, IIRC)...would I need more or would that do?

C

#8 lore raw
Quoted 5 lines · Christopher C. Chimelis

I think that's all we have, but I'm not sure. I'll look at the one that I do have and see what the "norm" is around the office (the one I have has to remain HP-UX for work reasons, but it should be a representative sample from what I've seen so far). These things also don't have much RAM (mine has 64MB, IIRC)...would I need more or would that do?

ok, just make sure you get a disc attached to the narrow interface, or you'll be limited to NFS... 64MB is fine.

Signature · 1 line
Revolutions do not require corporate support.
#9 lore raw
Quoted 2 lines · Matthew Wilcox

ok, just make sure you get a disc attached to the narrow interface, or you'll be limited to NFS... 64MB is fine.

Ok :-) I'll see what I can muster this week. We seem to be differential-poor drive-wise, so I don't think I'll have a problem getting a narrow drive :-) Also, we have a few external units laying around that are all narrow, so I may just use one of those.

C

#10 lore raw

I've posted debs for testing at http://gluck.debian.org/~tausq/hppa/

The following are available:

binutils (with latest patches from Alan Modra)
gcc (with ldcw alignment patch)
glibc (with latest patches from Alan Modra)
python and apache rebuilt with new toolchain bits

these are all unofficial.... Chris/Ben will be uploading official binutils/glibc. gcc will be uploaded as soon as the new binutils are in incoming....

there're still some problems with exception handling for g++, but in general these should work.... pls post to the list if you find otherwise.

randolph

Signature · 2 lines
Debian Developer <[email protected]>
http://www.TauSq.org/
#11 lore raw

Hello,

Quoted 8 lines · Randolph Chung

I've posted debs for testing at http://gluck.debian.org/~tausq/hppa/

The following are available:

binutils (with latest patches from Alan Modra) gcc (with ldcw alignment patch)
glibc (with latest patches from Alan Modra) python and apache rebuilt with new toolchain bits

<snippered>

I'm downloading these now, along with fresh debs. I am going to setup a nfsroot again to test. Hopefully I can finally run ls on my 720 now? :-)

I notice that the latest kernel on http://ftp.parisc.linux.org/ is linux-20011024.tar.gz. Are the snapshots lagging behind, or was this the last working snapshot? :-)

Thanks for the hard work!

Regards,

Albert

#12 lore raw

Hello,

Things are looking good. I've been running tar and ls and whatnot. Fired up screen to see what was going on. Seems I have a bit of a sick apt db, and /usr/sbin/install-info just sits there sometimes, but I think most of this can be attributed to a) nfsroot wierdness and b) a sick apt db.

I'll play around some more. Don't know when my local mirror is going to have these packages, so I'm going to try to setup a mirror on one of my machines again over the weekend and attempt a fresh install with the 3.0.15 boot-floppies.

Regards,

Albert

#13 lore raw

Hello,

I've managed to battle with apt a bit more. Some more stuff is now installed. I broke screen though. Some nfs thing.

Anyway, next mini-hurdle:

init-2.05# apt-get update
Get:1 http://ftp.leg.uct.ac.za testing/main Packages [1139kB] Get:Illegal instruction

dmesg output:

!!die_if_kernel: apt-get(3956): Privileged register - shouldn't happen! 11

YZrvWESTHLNXBCVMcbcbcbcbOGFRQPDI
PSW: 00000000000001101111111100001111

r0-3     00000000 401e43fc 401845ff 00045990
r4-7     401e5bfc faf00d10 faf00b90 0000000a
r8-11    0004d1b0 faf00948 0004d1d8 faf00ba8
r12-15   00000000 faf00b98 000c5210 00000000
r16-19   00000000 00000000 10348000 00045220
r20-23   00000002 401cec20 401845b0 00000000
r24-27   00000020 00045998 faf00d90 00040220
r28-31   faf00d10 00000005 faf00e00 00026f03
sr0-3    00000000 0000004c 00000000 0000004c
sr4-7    0000004c 0000004c 0000004c 0000004c

IASQ: 0000004c 0000004c IAOQ: 401cec23 401cec27
IIR: 00040948 ISR: 0000004c IOR: 00000000
CPU: 0 CR30: 15bb0000 CR31: 10350000
ORIG_R28: 00000001

Clearly it does happen. Mehehe. :-)

Should I decode this? I think I'll find instructions in the list archive? Search for "decode"?

That's about it for now. Keep on rollin'! :-)

Regards,

Albert

#14 lore raw

Hello,

Forgot to add the kernel info to previous post...

init-2.05# uname -a
Linux bob 2.4.9-pa70 #5 Tue Nov 6 12:58:22 SAST 2001 parisc unknown

This is the latest snapshot on ftp.parisc-linux.org.

Regards,

Albert

#15 lore raw
In reply to Albert Strasheim
Quoted 9 lines · Albert Strasheim

IASQ: 0000004c 0000004c IAOQ: 401cec23 401cec27
IIR: 00040948 ISR: 0000004c IOR: 00000000
CPU: 0 CR30: 15bb0000 CR31: 10350000
ORIG_R28: 00000001

Clearly it does happen. Mehehe. :-)

Should I decode this? I think I'll find instructions in the list archive? Search for "decode"?

401cec23 is in a library... it might need to get recompiled too.

Signature · 1 line
Revolutions do not require corporate support.
#16 lore raw
Quoted 25 lines · Matthew Wilcox
Quoted 18 lines · Albert Strasheim

YZrvWESTHLNXBCVMcbcbcbcbOGFRQPDI
PSW: 00000000000001101111111100001111

r0-3     00000000 401e43fc 401845ff 00045990
r4-7     401e5bfc faf00d10 faf00b90 0000000a
r8-11    0004d1b0 faf00948 0004d1d8 faf00ba8
r12-15   00000000 faf00b98 000c5210 00000000
r16-19   00000000 00000000 10348000 00045220
r20-23   00000002 401cec20 401845b0 00000000
r24-27   00000020 00045998 faf00d90 00040220
r28-31   faf00d10 00000005 faf00e00 00026f03
sr0-3    00000000 0000004c 00000000 0000004c
sr4-7    0000004c 0000004c 0000004c 0000004c

IASQ: 0000004c 0000004c IAOQ: 401cec23 401cec27
IIR: 00040948 ISR: 0000004c IOR: 00000000
CPU: 0 CR30: 15bb0000 CR31: 10350000
ORIG_R28: 00000001

Quoted 4 lines

Clearly it does happen. Mehehe. :-)

Should I decode this? I think I'll find instructions in the list archive? Search for "decode"?

401cec23 is in a library... it might need to get recompiled too.

#17 lore raw
Quoted 1 line · Richard Hirst

apt-get works fine for me, btw (715/75 and a500), running apt 0.5.4

Hmm, having switch from chroot to real root on the a500, I'm now seeing this same SIGILL.

Richard

#18 lore raw

Hello,

Seems with all the apt db madness, libstdc++ didn't get installed the first time. I installed it again, and things worked (better). There is something strange going on with my nfsroot. Don't know if it's some nfs setting or something... I get "too many symlinks" and "I/O" errors.

Quoted 5 lines · Richard Hirst
Quoted 1 line · Richard Hirst

apt-get works fine for me, btw (715/75 and a500), running apt 0.5.4

Hmm, having switch from chroot to real root on the a500, I'm now seeing this same SIGILL.

Hehe. Can't help you there...

Anyway, I'm going to attempt a 3.0.15 boot-floppies install now. I presume this is probably not going to work yet? We'll soon find out...

Seems the lifimage I crafted with palo is a bit ill. I'm going try to kick my nfsroot into shape, and try to debootstrap onto my SCSI disk.

Regards,

Albert

#19 lore raw

Hello,

Whoohooo. I bootstrapped my SCSI disk by hand (took a while; thank goodness for Perl and sed).

And now... it just works! :-) Almost.

I can boot single-user mode with /sbin/init, but if I boot multi-user (runlevel 2), everything starts fine up until cron comes up -- then minicom goes Offline and the system is MIA. I currently still have start-stop-daemon replaced with /bin/true, so I don't know why it's even doing much. Now that I think about it, it might be something that comes after cron. The getty perhaps? What needs to get configured here to make it work with serial console?

bash-2.05# uname -a
Linux bob 2.4.9-pa70 #13 Wed Nov 7 00:25:29 SAST 2001 parisc unknown

apt-get install works. ls works. tar works. lynx works. Gonna try gcc now. I think I can find some source to compile.

Thanks again Alan and Randolph. And thanks Matt. And thanks Debian. And thanks HP... muhehehehehe.

Regards,

Albert

#20 lore raw
Quoted 2 lines · Albert Strasheim

after cron. The getty perhaps? What needs to get configured here to make it work with serial console?

Enable the relevant line in /etc/inittab, something like

T0:23:respawn:/sbin/getty -L ttyS0 9600 vt100

and add ttyS0 to /etc/securetty to enable root logins.

Richard

#21 lore raw
In reply to Randolph Chung

Hi all,

GOOD NEWS
The new packages seem to solve my troubles with Expect. It seems to be
some good patches in this version.

Thanks to all guys that made them available, Thomas. ESIEE Team

#22 lore raw
Quoted 7 lines · Thomas Marteau

Hi all,

GOOD NEWS
The new packages seem to solve my troubles with Expect. It seems to be
some good patches in this version.

Quoted 7 lines

Thanks to all guys that made them available, Thomas. ESIEE Team

_______________________________________________
parisc-linux mailing list
[email protected]
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Is it now safe again to do a apt-get upgrade?

-- Nathan

------------------------------------------------------------
Nathan Neulinger EMail: [email protected]
University of Missouri - Rolla Phone: (573) 341-4841
Computing Services Fax: (573) 341-4216

#23 lore raw
Quoted 1 line

Is it now safe again to do a apt-get upgrade?

For the moment, the files are in gluck.debian.org/~tausq/hppa. But I think they will soon be available via apt-get. But, I will wait for doing upgrade because they are tested so not uploaded but I can be wrong.

Thomas.
ESIEE Team

#24 lore raw
Quoted 2 lines

For the moment, the files are in gluck.debian.org/~tausq/hppa. But I think they will soon be available via apt-get. But, I will wait for

Where can I find the sources+patches used ?

--
Kaj-Michael Lang , [email protected]

#25 lore raw
In reply to Thomas Marteau

FYI, I should have a new upload within the next day or two that includes the patches from Randolph and Alan. I'll let you know when the source package is in incoming so that you can build them and get them installed ASAP.

C

Quoted 10 lines · Thomas Marteau

Hi all,

GOOD NEWS
The new packages seem to solve my troubles with Expect. It seems to be
some good patches in this version.

Thanks to all guys that made them available, Thomas. ESIEE Team

#26 lore raw

[parisc-linux] Re: FW SCSI (Was hppa binutils/glibc updates)

In reply to Matthew Wilcox

What still needs to happen on FW diff SCSI to get it working?

Dan

Quoted 15 lines · Matthew Wilcox
Quoted 3 lines · Christopher C. Chimelis

If I think of something, I'll write ya. Also, if I manage to get a machine, I'm sure you'll be hearing from me (probably will be an Apollo series 735, fyi...slow, but effective).

735 is a pretty cool machine. don't get one which requires a fast-wide-diff scsi interface though; we don't have that working yet.

--
Revolutions do not require corporate support.

_______________________________________________
parisc-linux mailing list
[email protected]
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux