Re: [parisc-linux] gcc-3.2 experimental debs

Re: [parisc-linux] gcc-3.2 experimental debs

[Joel Soete]

Hi Randolph,

I try my hown gcc-3.2 (+ binutils, glibc) against kernel 2.4.18-pa19 
(not yet take the risc with pa20).

It compiles as well as with gcc3.0.x. It boots also but as soon as I try 
to connect it remotly with telnet or ssh (I forget to test ftp) the 
system crash:
"Stack pointer and cr30 do not correspond!
Dumping virtual address stack instead

Dumping Stack from ... to ...:

...

[those tree line seems to be repeated infinitely. well after 5 minutes I 
had to unplug the power supply to reboot]

Humm as it appends as well with telnet as ssh I suspect a inetd problem?

Any idea?

Thanks,
	Joel

PS: I will try your dpkg asap

Re: [parisc-linux] gcc-3.2 experimental debs

[Matthew Wilcox]

On Sun, Apr 21, 2002 at 01:05:59PM +0000, Joel Soete wrote:
> "Stack pointer and cr30 do not correspond!
> Dumping virtual address stack instead
> 
> Dumping Stack from ... to ...:
> 
> Humm as it appends as well with telnet as ssh I suspect a inetd problem?

I'd bet on gcc-3.2 miscompiling something in the network part of the kernel,
personally.

-- 
Revolutions do not require corporate support.

Re: [parisc-linux] gcc-3.2 experimental debs

[Carlos O'Donell Jr.]

> > "Stack pointer and cr30 do not correspond!
> > Dumping virtual address stack instead
> > 
> > Dumping Stack from ... to ...:
> > 
> > Humm as it appends as well with telnet as ssh I suspect a inetd problem?
> 
> I'd bet on gcc-3.2 miscompiling something in the network part of the kernel,
> personally.
>

Tested today.

Happens in the 3.1 main branch aswell (don't remember it previously happening).
Maybe I shouldn't be recommending the use of 3.1 as a cross compiler? ;)

I'll kdb on a serial tommorow morning and see what is causing it to die.

Oddly enough, I run diskless systems and they run off NFS perfectly fine...
so everything can't be that broke.

c. 

Re: [parisc-linux] gcc-3.2 experimental debs

[joel.soete]

Hello Carlos,

Quoting "Carlos O'Donell Jr." <carlos@baldric.uwo.ca>:

> > > "Stack pointer and cr30 do not correspond!
> > > Dumping virtual address stack instead
> > > 
> > > Dumping Stack from ... to ...:
> > > 
> > > Humm as it appends as well with telnet as ssh I suspect a inetd
> problem?
> > 
> > I'd bet on gcc-3.2 miscompiling something in the network part of the
> kernel,
> > personally.
> >
> 
> Tested today.
> 
> Happens in the 3.1 main branch aswell (don't remember it previously
> happening).
> Maybe I shouldn't be recommending the use of 3.1 as a cross compiler?
> ;)

I also notice it on 3.1 about two mount ago (that is why I do another try with 3.2)


> 
> I'll kdb on a serial tommorow morning and see what is causing it to
> die.
> 

I would also to do so. But I had the bad idea to co the last gcc release
(yesterday and today) to rebuild the toolchain.
But gcc build seems to be broken (step 4.2 of your procedure):

gcc -c -DIN_GCC    -g -O2 -W -Wall -Wwrite-strings -Wstrict-prototypes
-Wmissing-prototypes   -DHAVE_CONFIG_H    -I. -Ijava
-I/GNU-Dev/parisc-linux/src/gcc/gcc -I/GNU-Dev/parisc-linux/src/gcc/gcc/java
-I/GNU-Dev/parisc-linux/src/gcc/gcc/config
-I/GNU-Dev/parisc-linux/src/gcc/gcc/../include
/GNU-Dev/parisc-linux/src/gcc/gcc/java/parse.c -o java/parse.o
/tmp/ccj2Nlrs.s: Assembler messages:
/tmp/ccj2Nlrs.s:148746: Error: Field out of range [-262144..262143] (-263392).
make[1]: *** [java/parse.o] Error 1
make[1]: Leaving directory `/GNU-Dev/parisc-linux/build/gcc/gcc'
make: *** [all-gcc] Error 2

Last week co did not try to build any java.
Any idea or work-around?

Joel

PS1: about step 4.2 of your howto-xc, I notice a small difference between html
and pdf (in html you use gcc option "enable-shared" in pdf "disable-shared").
But this not the cause of the above problem.

PS2: I also make an apt-get dist-upgrade (with ftp.de.debian.org) so that native
gcc is now 3.0.4-5 (I think it was already installed last week)




-------------------------------------------------
This mail sent through Tiscali Webmail (http://webmail.tiscali.be)

Re: [parisc-linux] gcc-3.2 experimental debs

[John David Anglin]

> I would also to do so. But I had the bad idea to co the last gcc release
> (yesterday and today) to rebuild the toolchain.
> But gcc build seems to be broken (step 4.2 of your procedure):
> 
> gcc -c -DIN_GCC    -g -O2 -W -Wall -Wwrite-strings -Wstrict-prototypes
> -Wmissing-prototypes   -DHAVE_CONFIG_H    -I. -Ijava
> -I/GNU-Dev/parisc-linux/src/gcc/gcc -I/GNU-Dev/parisc-linux/src/gcc/gcc/java
> -I/GNU-Dev/parisc-linux/src/gcc/gcc/config
> -I/GNU-Dev/parisc-linux/src/gcc/gcc/../include
> /GNU-Dev/parisc-linux/src/gcc/gcc/java/parse.c -o java/parse.o
> /tmp/ccj2Nlrs.s: Assembler messages:
> /tmp/ccj2Nlrs.s:148746: Error: Field out of range [-262144..262143] (-263392).
> make[1]: *** [java/parse.o] Error 1
> make[1]: Leaving directory `/GNU-Dev/parisc-linux/build/gcc/gcc'
> make: *** [all-gcc] Error 2
> 
> Last week co did not try to build any java.
> Any idea or work-around?

"make bootstrap".

Dave
-- 
J. David Anglin                                  dave.anglin@nrc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)

Re: [parisc-linux] gcc-3.2 experimental debs

[Carlos O'Donell Jr.]

> 
> Joel
> 
> PS1: about step 4.2 of your howto-xc, I notice a small difference between html
> and pdf (in html you use gcc option "enable-shared" in pdf "disable-shared").
> But this not the cause of the above problem.
> 
> PS2: I also make an apt-get dist-upgrade (with ftp.de.debian.org) so that native
> gcc is now 3.0.4-5 (I think it was already installed last week)
> 
>

Both the HTML and PDF version are the same (just double checked).

http://www.baldric.uwo.ca/~carlos/HOWTO-Parisc/index.html
http://www.baldric.uwo.ca/~carlos/HOWTO-Parisc/HOWTO-Building_PARISC-LINUX_CrossCompiler.pdf

In section 4.2 you are bootstraping the C compiler without glibc
so you must use --disable-shared.

In section 4.5 you have already built glibc and you can turn
shared support back on with --enable-shared.

c.

p.s. The HTML and PDF do have some formatting differences...
     I'll get around to making them look similar later :}
     Dang... just noticed I missed one of the path changes in
     the pdf...

Re: [parisc-linux] gcc-3.2 experimental debs

[joel.soete]

Quoting "Carlos O'Donell Jr." <carlos@baldric.uwo.ca>:

> > 
> > Joel
> > 
> > PS1: about step 4.2 of your howto-xc, I notice a small difference
> between html
> > and pdf (in html you use gcc option "enable-shared" in pdf
> "disable-shared").
> > But this not the cause of the above problem.
> > 
> > PS2: I also make an apt-get dist-upgrade (with ftp.de.debian.org) so
> that native
> > gcc is now 3.0.4-5 (I think it was already installed last week)
> > 
> >
> 
> Both the HTML and PDF version are the same (just double checked).
> 
> http://www.baldric.uwo.ca/~carlos/HOWTO-Parisc/index.html
>
http://www.baldric.uwo.ca/~carlos/HOWTO-Parisc/HOWTO-Building_PARISC-LINUX_CrossCompiler.pdf
> 
> In section 4.2 you are bootstraping the C compiler without glibc
> so you must use --disable-shared.
> 
> In section 4.5 you have already built glibc and you can turn
> shared support back on with --enable-shared.
> 
> c.

Yes you have right (as it is my html which was wrong I do have forgot to
'reload' the page). Sorry for confusion.

Thanks also for explanation and attention.
    Joel

> 
> p.s. The HTML and PDF do have some formatting differences...
>      I'll get around to making them look similar later :}
>      Dang... just noticed I missed one of the path changes in
>      the pdf...
> 
> 


-------------------------------------------------
This mail sent through Tiscali Webmail (http://webmail.tiscali.be)

Re: [parisc-linux] gcc-3.2 experimental debs

[joel.soete]

Quoting joel.soete@freebel.net:

> 
> gcc -c -DIN_GCC    -g -O2 -W -Wall -Wwrite-strings -Wstrict-prototypes
> -Wmissing-prototypes   -DHAVE_CONFIG_H    -I. -Ijava
> -I/GNU-Dev/parisc-linux/src/gcc/gcc
> -I/GNU-Dev/parisc-linux/src/gcc/gcc/java
> -I/GNU-Dev/parisc-linux/src/gcc/gcc/config
> -I/GNU-Dev/parisc-linux/src/gcc/gcc/../include
> /GNU-Dev/parisc-linux/src/gcc/gcc/java/parse.c -o java/parse.o
> /tmp/ccj2Nlrs.s: Assembler messages:
> /tmp/ccj2Nlrs.s:148746: Error: Field out of range [-262144..262143]
> (-263392).
> make[1]: *** [java/parse.o] Error 1
> make[1]: Leaving directory `/GNU-Dev/parisc-linux/build/gcc/gcc'
> make: *** [all-gcc] Error 2


Forget. The actual problem was again typo: in configure step I type
'enable-langages=c' in place of 'enable-languages=c' (lack of a u between g & a).

Sorry for mistakes.
    Joel


-------------------------------------------------
This mail sent through Tiscali Webmail (http://webmail.tiscali.be)

Re: [parisc-linux] gcc-3.2 experimental debs

[Randolph Chung]

> I try my hown gcc-3.2 (+ binutils, glibc) against kernel 2.4.18-pa19 
> (not yet take the risc with pa20).
> 
> It compiles as well as with gcc3.0.x. It boots also but as soon as I try 
> to connect it remotly with telnet or ssh (I forget to test ftp) the 
> system crash:

if you still feel adventurous, check out the kdb cvs branch of the
kernel and you'll at least be able to tell where it crashed... :-)

randolph

Re: [parisc-linux] gcc-3.2 experimental debs

[joel.soete]

Quoting Randolph Chung <tausq@debian.org>:

> > I try my hown gcc-3.2 (+ binutils, glibc) against kernel 2.4.18-pa19 
> > (not yet take the risc with pa20).
> > 
> > It compiles as well as with gcc3.0.x. It boots also but as soon as I
> try 
> > to connect it remotly with telnet or ssh (I forget to test ftp) the 
> > system crash:
> 
> if you still feel adventurous, check out the kdb cvs branch of the
> kernel and you'll at least be able to tell where it crashed... :-)
> 
> randolph
> 
> 

Yes I will also try (and find out howto use)

Joel

-------------------------------------------------
This mail sent through Tiscali Webmail (http://webmail.tiscali.be)

[parisc-linux] gcc-3.2 experimental debs

[Randolph Chung]

I've uploaded an experimental gcc-3.2 debian package to:

ftp://ftp.parisc-linux.org/unofficial-debs/gcc-3.2_20020417-1_hppa.deb 

for people who feel adventurous and wants to try. As the name suggests
it's built from gcc-3.2 cvs from two days ago. The deb is unfortunately
quite large (26M) because it also contains binutils bits and has
c/c++/fortran/objc compilers enabled.

The regression test results are looking quite good, and at least on the
two cases I've tried the functions-sections problems are fixed.

The package installs into /usr/local, and I want to stress here
that these are only for experimental use and are not part of the
official debian release.

Having said that, try them out and let us know how it works out!

Last but not least, thanks Dave for all your efforts! gcc-3.2 looks
really quite a bit better than 3.0.x

randolph