Looney Toon tries his hand at something slightly more complex thanstraight Fortran 77

Looney Toon tries his hand at something slightly more complex thanstraight Fortran 77

[Toon Moene]

Bear with me ;-)

Courtesy of Greg Lindahl, I've access to an Alpha/Linux (Red Hat  
4.2) 500 Mhz, 21164 box.  config.guess says:

alphaev5-unknown-linux-gnulibc1

I've set up the CVS connection, downloaded a fresh copy this  
afternoon and started building.  Because I am a nobody on this  
machine, I --prefix everything with $HOME/usrlocal <grin>

So far, so good.  However, when trying to link f771 (The Holy Grail  
of this endeavour), the build dies with the dreaded "Truncated  
relocation to fit: blah-blah" message.

Because I tend to keep up with the egcs and egcs-bugs mailing  
lists, I thought I knew what the solution was (it is, however, not  
mentioned in the FAQ):  Get a recent version of binutils.

So I did.  Note that the link to binutils (source) in the FAQ is  
stale (.20 - it's now at .23).

Built binutils 2.8.1.0.23 with --prefix as above.  No complaints,  
except for some scary-looking "Warning: cast from pointer to integer  
of different size" messages <brrr>

Emptied build directory of egcs.  Reconfigured (same --prefix).   
Typed `make bootstrap' again.

Build dies when trying to execute ./gengenrtl.  Gdb shows:

$ gdb gcc/gengenrtl
GDB is free software and you are welcome to distribute copies of it
 under certain conditions; type "show copying" to see the conditions.
There is absolutely no warranty for GDB; type "show warranty" for  
details.
GDB 4.16 (alpha-unknown-linux), Copyright 1996 Free Software  
Foundation, Inc...
(gdb) run
Starting program: /part2/burley/obj/gcc/gengenrtl

Program received signal SIGSEGV, Segmentation fault.
0x1555555e3e8 in fixup ()
(gdb) bt
#0  0x1555555e3e8 in fixup ()
#1  0x1555555e688 in _dl_runtime_resolve ()
(gdb) quit

What does this mean (if anything) ?

Do I also need to upgrade to glibc-2 (shudder) ?  How does one do  
that, as a nobody ?

[ And then to recall that this all started as an effort to check
  Richard Henderson's "negative array bound" patch, sigh ]

Confusingly yours,
Toon.

Re: Looney Toon tries his hand at something slightly more complex than straight Fortran 77

[Richard Henderson]

> Program received signal SIGSEGV, Segmentation fault.
> 0x1555555e3e8 in fixup ()
> (gdb) bt
> #0  0x1555555e3e8 in fixup ()
> #1  0x1555555e688 in _dl_runtime_resolve ()
> (gdb) quit
> 
> What does this mean (if anything) ?

It means that, due to bugs in older versions of binutils, it is not as
simple as it ought to be to upgrade from 2.7.0.2 (shipped with RH4.x)
to 2.8.1.x.

The best solution is to upgrade to RH 5.0.  If it is absolutely required
that the machine stay at RH 4.2, I may be able to provide replacement RPMs.

> Do I also need to upgrade to glibc-2 (shudder) ?  How does one do  
> that, as a nobody ?

You don't.  You have to be root to upgrade.


r~

Re: Looney Toon tries his hand at something slightly more complexthan straight Fortran 77

[Toon Moene]

[ ... describing failure when trying to combine egcs-980402-CVS with
      binutils-2.8.1.0.23 ... ]

I wrote:

>> Program received signal SIGSEGV, Segmentation fault.
>> 0x1555555e3e8 in fixup ()
>> (gdb) bt
>> #0  0x1555555e3e8 in fixup ()
>> #1  0x1555555e688 in _dl_runtime_resolve ()
> (gdb) quit
>>
>> What does this mean (if anything) ?

Richard Henderson replied:

>  It means that, due to bugs in older versions of binutils,
>  it is not as simple as it ought to be to upgrade from
>  2.7.0.2 (shipped with RH4.x) to 2.8.1.x.
>
>  The best solution is to upgrade to RH 5.0.  If it is
>  absolutely required that the machine stay at RH 4.2, I
>  may be able to provide replacement RPMs.

Hmmm, as I noted before, this isn't *my* machine ... could you name  
the RPMs (and whether they can be installed by Someone Of Little  
Privilege) ?

>> Do I also need to upgrade to glibc-2 (shudder) ?  How does one do  
>> that, as a nobody ?

> You don't.  You have to be root to upgrade.

OK, that's out - then.

Thanks,
Toon.

Re: Looney Toon tries his hand at something slightly more complex than straight Fortran 77

[Harvey J. Stein]

Richard Henderson <rth@cygnus.com> writes:

> It means that, due to bugs in older versions of binutils, it is not as
> simple as it ought to be to upgrade from 2.7.0.2 (shipped with RH4.x)
> to 2.8.1.x.
> 
> The best solution is to upgrade to RH 5.0.  If it is absolutely required
> that the machine stay at RH 4.2, I may be able to provide replacement RPMs.

I'd like a replacement RPM, please.  I have a bunch of machines
running RH 4.2, and I can't upgrade to RH 5.0 due to various bugs &
problems with it.  And, I'm seeing quite a few programs that I can't
build because of an old binutils.

Thanks,

Harvey J. Stein
Berger Financial Research
hjstein@bfr.co.il