headers
matthew green | 1 Sep 2003 03:22
Picon
Favicon

re: GCC3.3.1 switch coming soon.


   >
   > i'm ready to switch sparc, sparc64, i386 & alpha ports to using GCC3.3.1
   > by default.  i've uploaded 4 snapshots (one per port ;-), all cross
   > compiled from i386-netbsd.  you can find them under
   > ftp.netbsd.org:/pub/NetBSD/arch/≤arch>/snapshot/snapshot-20030828-GCC3.1.1
   >

   does kernels compile well??

   while i already have 3-weeks old -current compiled with 3.3.1 with pentium
   optimization, i use 2.95.* from pkgsrc to compile kernels.
   kernels compiled with 3.3.* just reboots after loaded.

i've been running gcc3 compiled everything on my new laptop for several
weeks now.  the only time it had gcc2 was after installing and before
bootstraping with gcc3.  i've seen the reports about topdown VM but i
have not investigated them.  someone who uses that option should, i
have enough other things on my plate right now.  my laptop kernel is
built with:

	makeoptions     DEBUG="-gstabs" # compile full symbol table
	makeoptions     COPTS="-O4 -mcpu=pentium4 -march=pentium4"

.mrg.
Permalink | Reply | 0 comments |
headers
enami tsugutomo | 1 Sep 2003 15:09
Picon

re: GCC3.3.1 switch coming soon.

> i've seen the reports about topdown VM but i have not investigated
> them.  someone who uses that option should, i have enough other
> things on my plate right now.

It isn't compiler matter since the kernel built by 2.95.3 as well.  It
is actually a bug in topdown vm code.  It incorrectly re-uses space
already in-use under some condition which began to met recently.

Attached is a change I'm currently playing with.  It includes a fix
among other cleanups.

enami.

Index: uvm_map.c
===================================================================
RCS file: /cvsroot/src/sys/uvm/uvm_map.c,v
retrieving revision 1.137
diff -c -r1.137 uvm_map.c
*** uvm_map.c	26 Aug 2003 15:12:18 -0000	1.137
--- uvm_map.c	1 Sep 2003 07:54:05 -0000
***************
*** 199,204 ****
--- 199,206 ----
  static void uvm_map_entry_unwire __P((struct vm_map *, struct vm_map_entry *));
  static void uvm_map_reference_amap __P((struct vm_map_entry *, int));
  static void uvm_map_unreference_amap __P((struct vm_map_entry *, int));
+ static int uvm_map_space_avail __P((vaddr_t *, vsize_t, voff_t, vsize_t, int,
+ 	       struct vm_map_entry *));

  /*
(Continue reading)

Permalink | Reply | 0 comments |
headers
Jason Thorpe | 2 Sep 2003 16:34

Re: GCC3.3.1 switch coming soon.


On Sunday, August 31, 2003, at 02:30  AM, matthew green wrote:

> i'm ready to switch sparc, sparc64, i386 & alpha ports to using  
> GCC3.3.1
> by default.  i've uploaded 4 snapshots (one per port ;-), all cross
> compiled from i386-netbsd.  you can find them under
> ftp.netbsd.org:/pub/NetBSD/arch/≤arch>/snapshot/snapshot-20030828- 
> GCC3.1.1

Please note that GCC 3.3.1 creates debug info (DWARF2) that our current  
5.0-based GDB cannot understand.  Therefore, we need to either tweak  
GCC 3.3.1 to use DBX debug or throw the switch on GDB for those ports,  
as well.

         -- Jason R. Thorpe <thorpej <at> wasabisystems.com>
Permalink | Reply | 0 comments |
headers
Andrew Brown | 3 Sep 2003 03:31

Re: GCC3.3.1 switch coming soon.

On Mon, Sep 01, 2003 at 10:09:29PM +0900, enami tsugutomo wrote:
>> i've seen the reports about topdown VM but i have not investigated
>> them.  someone who uses that option should, i have enough other
>> things on my plate right now.
>
>It isn't compiler matter since the kernel built by 2.95.3 as well.  It
>is actually a bug in topdown vm code.  It incorrectly re-uses space
>already in-use under some condition which began to met recently.

where?  tell me tell me.  :)

>Attached is a change I'm currently playing with.  It includes a fix
>among other cleanups.

does the patch you sent rely on other recent changes?  i tried
applying it to my 8/10 source tree and while the kernel compiled
happily, it panicked very early on...early enough that it would not
generate a kernel dump.

it also looks like the patch drops some of the required changes to
deal with topdown...

--

-- 
|-----< "CODE WARRIOR" >-----|
codewarrior <at> daemon.org             * "ah!  i see you have the internet
twofsonet <at> graffiti.com (Andrew Brown)                that goes *ping*!"
werdna <at> squooshy.com       * "information is power -- share the wealth."
Permalink | Reply | 0 comments |
headers
Daniel Carosone | 3 Sep 2003 10:31
Picon

alignment fault crashes early in boot

It's early in testing yet, but I just converted an ultra60
to HAVE_GCC3=yes kernel and userland (pkgsrc building now).

So far, it looks good except for a mysterious problem at boot.

Sometimes on booting the new kernel, it will panic very early in
the boot process (just after copyright), typically in pool_init,
with an alignment fault.  These happen fairly often (5 out of 8
boots, so far).

The panic is an alignment fault.

There's not much of interest in the traceback at this point (main),
but just ask for other info that might be useful.  I didn't have
it on serial console at the time, but that can be arranged easily
enough.

If the boot survives past this point, it seems to be stable, though
I haven't really put it under load yet. It's built a dozen or so
of the usual suspects from pkgsrc so far, and I'll leave it rebuilding
src again overnight.

--
Dan.
Permalink | Reply | 1 comment |
headers
Martin Husemann | 3 Sep 2003 16:13
Picon

Re: alignment fault crashes early in boot

On Wed, Sep 03, 2003 at 06:31:02PM +1000, Daniel Carosone wrote:
> So far, it looks good except for a mysterious problem at boot.

Did you install the new bootblocks too? I didn't and I've been running
gcc 3.3.1 compiled kernel and userland on two machines for quite some
time now without any problems.

Martin
Permalink | Reply | 0 comments |
headers
Alexander Zich | 4 Sep 2003 18:43
Picon

RE: libiconv 1.9.1 on NetBSD 1.6.1 (port sparc64)

Hey nice,

you found my message :-)

Well I can tell you how I got around the problem:

When you have a look to error.h line 50 where the error funtions are
defined. They are enclosed in an #ifdef block. Just force the compiler to
use the #elseif branch (for example by adding a 2 to the #ifdef line).

You will then get another error in unsetenv.c. I think you can get around
that by changing the return type of unsetenv() from int to void. You will
get warnings but no errors.

Try a bit, it should work. Please note this is a hack. could compile it and
'make test' also works but there might be "side effects".

Hope this could help.

Bye,
  Alex

-----Original Message-----
From: port-sparc64-owner <at> NetBSD.org
[mailto:port-sparc64-owner <at> NetBSD.org]On Behalf Of Dario Borchers
Sent: Tuesday, May 13, 2003 4:26 PM
To: port-sparc64 <at> netbsd.org
Subject: libiconv 1.9.1 on NetBSD 1.6.1 (port sparc64)

Hi,
(Continue reading)

Permalink | Reply | 0 comments |
headers
leam | 5 Sep 2003 02:46
Favicon

Failure to build gcc3 on sparc64

In /usr/pkgsrc/lang/gcc3 I'm trying to build gcc3 and getting:

# make
make: "/usr/pkgsrc/lang/gcc3/../../mk/bsd.pkg.defaults.mk" line 88: Need 
an operator
make: "/usr/pkgsrc/lang/gcc3/../../mk/bsd.pkg.defaults.mk" line 90: 
Makefile appears to contain unresolved cvs/rcs/??? merge conflicts
make: "/usr/pkgsrc/lang/gcc3/../../mk/bsd.pkg.defaults.mk" line 104: 
Need an operator
make: "/usr/pkgsrc/lang/gcc3/../../mk/bsd.pkg.defaults.mk" line 160: 
Need an operator
make: "/usr/pkgsrc/lang/gcc3/../../mk/bsd.pkg.defaults.mk" line 162: 
Makefile appears to contain unresolved cvs/rcs/??? merge conflicts
make: "/usr/pkgsrc/lang/gcc3/../../mk/bsd.pkg.defaults.mk" line 174: 
Need an operator
make: Fatal errors encountered -- cannot continue

make: stopped in /usr/pkgsrc/lang/gcc3

###

The bsd.pkg.defaults.mk lines in question have things like "<<<<<<" and 
"=====". Some time has past since my last build, what am I forgetting? 
What do I need to re-read?

ciao!

leam
(Continue reading)

Permalink | Reply | 5 comments |
headers
Greywolf | 5 Sep 2003 03:05

Re: Failure to build gcc3 on sparc64

Thus spake leam ("l> ") sometime Today...

l> The bsd.pkg.defaults.mk lines in question have things like "<<<<<<" and
l> "=====". Some time has past since my last build, what am I forgetting?
l> What do I need to re-read?

You need to rm bsd.pkg.defaults.mk and refetch it.

[that's the simplest way...]

				--*greywolf;
--
NetBSD: Stop, Drop, and Load
Permalink | Reply | 0 comments |
headers
Hubert Feyrer | 5 Sep 2003 03:23
Picon

Re: Failure to build gcc3 on sparc64

On Thu, 4 Sep 2003, leam wrote:
> make: "/usr/pkgsrc/lang/gcc3/../../mk/bsd.pkg.defaults.mk" line 90:
> Makefile appears to contain unresolved cvs/rcs/??? merge conflicts
...
> What do I need to re-read?

The above line, then clean up the mess. :)

 - Hubert

--

-- 
  ___ _ _  _   _        * Harddisk Image Cloning *
 / __| | || | | |           www.feyrer.de/g4u/
| (_ |_  _| |_| |
 \___| |_| \___/          v1.12 out now, including partition support!
Permalink | Reply | 3 comments |

Gmane