Re: gcc-4.6.0 breaks

Re: gcc-4.6.0 breaks

[Dennis Clarke]

> Dennis Clarke wrote:
>>>     My gcc-4.6.0 compilation after compiling for quite some time
>>> decided to
>>> break with an error to stderr of something like not recognized object
>>> files.
>>> I am compiling with gcc-3.3.6. This can sure be nerve racking. Does
>>> anyone know what this means?
>>>
>>
>> How would anyone even begin to help with this email content ?
>>
>> Okay, so let's see, so you're running Debian Linux on a DEC alpha and
>> you are trying to bootstrap GCC 4.6.0 while using GCC 3.3.6 with
>> binutils maybe maybe not and your configure line was black magic
>> while facing east with a dead cat under a rising moon ?
>>
>> My guess is that you need to face west.
>>
>> Try again with some content please !
>
>     I didn't get a log of stderr. I wish I had it turned on now. I'm using
> an old RH 9 for the i686 and binutils-2.21 along with gcc-3.3.6. I have
> have
> gmp mpfc and mpc all installed and I am using make-3.82. I was rather
> surprised when it broke. If I try again I will get a stderr log.
>

Can you just login remotely with an xterm or better yet fire up screen
with session logging and then go to a fresh clean directory. Then let's
try a boot strap again and this time everything gets logged.

wdyt ?

Dennis

Re: gcc-4.6.0 breaks

[Bill Cunningham]

Dennis Clarke wrote:
>> Dennis Clarke wrote:
>>>>     My gcc-4.6.0 compilation after compiling for quite some time
>>>> decided to
>>>> break with an error to stderr of something like not recognized
>>>> object files.
>>>> I am compiling with gcc-3.3.6. This can sure be nerve racking. Does
>>>> anyone know what this means?
>>>>
>>>
>>> How would anyone even begin to help with this email content ?
>>>
>>> Okay, so let's see, so you're running Debian Linux on a DEC alpha
>>> and you are trying to bootstrap GCC 4.6.0 while using GCC 3.3.6 with
>>> binutils maybe maybe not and your configure line was black magic
>>> while facing east with a dead cat under a rising moon ?
>>>
>>> My guess is that you need to face west.
>>>
>>> Try again with some content please !
>>
>>     I didn't get a log of stderr. I wish I had it turned on now. I'm
>> using an old RH 9 for the i686 and binutils-2.21 along with
>> gcc-3.3.6. I have have
>> gmp mpfc and mpc all installed and I am using make-3.82. I was rather
>> surprised when it broke. If I try again I will get a stderr log.
>>
>
> Can you just login remotely with an xterm or better yet fire up screen
> with session logging and then go to a fresh clean directory. Then
> let's try a boot strap again and this time everything gets logged.
>
> wdyt ?

    Nope. Don't have X on my system. Never installed it and removed what few 
directories were left. I have xfree86 binaries but they are kind of old and 
want an even newer glibc than I have I believe. But I can still get errors 
logged with the shell.

Bill

Re: gcc-4.6.0 breaks

[Bryan Hundven]

On Wed, Jun 1, 2011 at 7:46 PM, Bill Cunningham <billcun@suddenlink.net> wrote:
> Dennis Clarke wrote:
>>>
>>> Dennis Clarke wrote:
>>>>>
>>>>>    My gcc-4.6.0 compilation after compiling for quite some time
>>>>> decided to
>>>>> break with an error to stderr of something like not recognized
>>>>> object files.
>>>>> I am compiling with gcc-3.3.6. This can sure be nerve racking. Does
>>>>> anyone know what this means?
>>>>>
>>>>
>>>> How would anyone even begin to help with this email content ?
>>>>
>>>> Okay, so let's see, so you're running Debian Linux on a DEC alpha
>>>> and you are trying to bootstrap GCC 4.6.0 while using GCC 3.3.6 with
>>>> binutils maybe maybe not and your configure line was black magic
>>>> while facing east with a dead cat under a rising moon ?
>>>>
>>>> My guess is that you need to face west.
>>>>
>>>> Try again with some content please !
>>>
>>>    I didn't get a log of stderr. I wish I had it turned on now. I'm
>>> using an old RH 9 for the i686 and binutils-2.21 along with
>>> gcc-3.3.6. I have have
>>> gmp mpfc and mpc all installed and I am using make-3.82. I was rather
>>> surprised when it broke. If I try again I will get a stderr log.
>>>
>>
>> Can you just login remotely with an xterm or better yet fire up screen
>> with session logging and then go to a fresh clean directory. Then
>> let's try a boot strap again and this time everything gets logged.
>>
>> wdyt ?
>
>   Nope. Don't have X on my system. Never installed it and removed what few
> directories were left. I have xfree86 binaries but they are kind of old and
> want an even newer glibc than I have I believe. But I can still get errors
> logged with the shell.

$ apt-get install screen

$ screen -L

$ build your toolchain

$ exit

post screenlog.0

> Bill
>
>

-Bryan

Re: gcc-4.6.0 breaks

[Miles Bader]

Bryan Hundven <bryanhundven@gmail.com> writes:
> $ apt-get install screen

Amen.

-miles

-- 
Cat, n. A soft, indestructible automaton provided by nature to be kicked when
things go wrong in the domestic circle.

Re: gcc-4.6.0 breaks

[Bryan Hundven]

On Wed, Jun 1, 2011 at 9:59 PM, Miles Bader <miles@gnu.org> wrote:
> Bryan Hundven <bryanhundven@gmail.com> writes:
>> $ apt-get install screen
>
> Amen.

HAHA, I just realized Bill said rh, not deb :-P

I like Jonathan's pencil idea too!
The manual printer, ftw!

$ yum install screen

> -miles
>
> --
> Cat, n. A soft, indestructible automaton provided by nature to be kicked when
> things go wrong in the domestic circle.

-Bryan

Re: gcc-4.6.0 breaks

[Bill Cunningham]

Jonathan Wakely wrote:
> On 2 June 2011 13:17, Bill Cunningham wrote:
>>
>> Ok but I have another question concerning gcc. I would like to have 2
>> compilers. A default and one perhaps in /usr/local or another
>> directory. How can I switch back and forth between which compiler I
>> want?
>
> The same way you'd choose between any two programs in different
> locations.
>
> You could adjust your $PATH or write a script which calls the one you
> want.
>
> I have a handful of versions installed in ~/gcc/ and invoke them via a
> script like
>
> #!/bin/bash
>> ${ver:=4.7}
>> ${libdir:=lib64}
> $HOME/gcc/$ver/bin/g++ -Wl,-rpath,$HOME/gcc/$ver/$libdir -Wall -g "$@"

    I see in .bash_profile that there is a PATH variable set. I have to put 
all my binaries in /usr to get the shell to recognize it. I know nothing 
about bash and would like to run a script that would allow me to simply run 
a script that would switch back and forth between compilers, assemblers, and 
the linker. Maybe 2 scripts. But I can't find the file that lists all my 
path data. env shows the path as does echo $PATH. I always log in as root 
anyway. If I want single user mode I add init=/bin/bash to the kernel switch 
of grub.

Bill

Re: gcc-4.6.0 breaks

[kevin diggs]

Hi,

On Wed, Jun 1, 2011 at 9:16 PM, Dennis Clarke <dclarke@blastwave.org> wrote:
>
>>     My gcc-4.6.0 compilation after compiling for quite some time decided
>> to
>> break with an error to stderr of something like not recognized object
>> files.
>> I am compiling with gcc-3.3.6. This can sure be nerve racking. Does anyone
>> know what this means?
>>
>
> How would anyone even begin to help with this email content ?
>
> Okay, so let's see, so you're running Debian Linux on a DEC alpha and you
> are trying to bootstrap GCC 4.6.0 while using GCC 3.3.6 with binutils
> maybe maybe not and your configure line was black magic while facing east
> with a dead cat under a rising moon ?
>
I thought it was "widely known" that under a rising moon, the cat HAD
to at least be on life support. It could not be completely dead?!?

> My guess is that you need to face west.
>
> Try again with some content please !
>
> Dennis
>
>
>

kevin

Re: gcc-4.6.0 breaks

[Jonathan Wakely]

On 2 June 2011 15:20, Marc Glisse wrote:
> On Thu, 2 Jun 2011, Jonathan Wakely wrote:
>
>> I have a handful of versions installed in ~/gcc/ and invoke them via a
>> script like
>>
>> #!/bin/bash
>> : ${ver:=4.7}
>> : ${libdir:=lib64}
>> $HOME/gcc/$ver/bin/g++ -Wl,-rpath,$HOME/gcc/$ver/$libdir -Wall -g "$@"
>
> When you have many versions, it may be more convenient to add a spec file
> that does the rpath thing. This way you can call any g++ in any way without
> bothering with wrappers and without risk.

I often don't want to use -rpath at all.  Now and then I need to
compile with one version and run against a different libstdc++.so to
ensure I haven't broken the ABI, so I mix and match g++ and
libstdc++.so versions.  But that's not something most people will ever
need to do. So optional wrapper scripts work for me.  YMMV.

Re: gcc-4.6.0 breaks

[Marc Glisse]

On Thu, 2 Jun 2011, Jonathan Wakely wrote:

> I have a handful of versions installed in ~/gcc/ and invoke them via a
> script like
>
> #!/bin/bash
> : ${ver:=4.7}
> : ${libdir:=lib64}
> $HOME/gcc/$ver/bin/g++ -Wl,-rpath,$HOME/gcc/$ver/$libdir -Wall -g "$@"

When you have many versions, it may be more convenient to add a spec file 
that does the rpath thing. This way you can call any g++ in any way 
without bothering with wrappers and without risk.

-- 
Marc Glisse

Re: gcc-4.6.0 breaks

[Jonathan Wakely]

On 2 June 2011 13:17, Bill Cunningham wrote:
>
>   Ok but I have another question concerning gcc. I would like to have 2
> compilers. A default and one perhaps in /usr/local or another directory. How
> can I switch back and forth between which compiler I want?

The same way you'd choose between any two programs in different locations.

You could adjust your $PATH or write a script which calls the one you want.

I have a handful of versions installed in ~/gcc/ and invoke them via a
script like

#!/bin/bash
: ${ver:=4.7}
: ${libdir:=lib64}
$HOME/gcc/$ver/bin/g++ -Wl,-rpath,$HOME/gcc/$ver/$libdir -Wall -g "$@"

Re: gcc-4.6.0 breaks

[Bill Cunningham]

Jonathan Wakely wrote:
> On 2 June 2011 12:21, Bill Cunningham wrote:
>> Jonathan Wakely wrote:
>>
>>>> No. Although I didn't manually use the switches with gcc to find
>>>> gmp mpc or mpfc. The configure script seemed to me atleast to be
>>>> able to find those libraries I put in /usr/local. No when
>>>> compiling mpc I believe it was it needed to know where gmp was.
>>>> The script needed help there. gmp mpc and mpfc were the only
>>>> things gcc to my knowledge needed to compile and run.
>>>
>>> There's a difference between configure being able to find the
>>> libraries to link an executable and the dynamic linker being able to
>>> find those libraris to *run* an executable.
>>>
>>> If /usr/local/lib is not in your ldconfig cache then it will not
>>> find libgmp.so etc. so you will need to reconfigure ldconfig, or set
>>> LD_LIBRARY_PATH.
>>>
>>> Of course since you don't know what the actual error was this is all
>>> just wild speculation. You need to know what the error was and if it
>>> told you to look in config.log, which would have more information.
>>>
>>> Try looking in the build directory for $target/libgcc/config.log
>>> which I suspect will show that ld.so couldn't find one of libgmp.so,
>>> libmpfr.so or libmpc.so
>>
>> Isn't the dynamic linker, ld.so and ld-linux.so2 the same thing?
>
> Yes.
>
> But ld is not the same thing as ld.so and my point is that ld might
> have been able to find the libraries, but ld.so could not.
>
> But this still seems like a waste of time until you know what the
> actual error was.  I still suggest looking in
> $target/libgcc/config.log to see if it's in there.

    Ok but I have another question concerning gcc. I would like to have 2 
compilers. A default and one perhaps in /usr/local or another directory. How 
can I switch back and forth between which compiler I want?

Bill

Re: gcc-4.6.0 breaks

[Jonathan Wakely]

On 2 June 2011 12:21, Bill Cunningham wrote:
> Jonathan Wakely wrote:
>
>>> No. Although I didn't manually use the switches with gcc to find gmp
>>> mpc or mpfc. The configure script seemed to me atleast to be able to
>>> find those libraries I put in /usr/local. No when compiling mpc I
>>> believe it was it needed to know where gmp was. The script needed
>>> help there. gmp mpc and mpfc were the only things gcc to my
>>> knowledge needed to compile and run.
>>
>> There's a difference between configure being able to find the
>> libraries to link an executable and the dynamic linker being able to
>> find those libraris to *run* an executable.
>>
>> If /usr/local/lib is not in your ldconfig cache then it will not find
>> libgmp.so etc. so you will need to reconfigure ldconfig, or set
>> LD_LIBRARY_PATH.
>>
>> Of course since you don't know what the actual error was this is all
>> just wild speculation.  You need to know what the error was and if it
>> told you to look in config.log, which would have more information.
>>
>> Try looking in the build directory for $target/libgcc/config.log which
>> I suspect will show that ld.so couldn't find one of libgmp.so,
>> libmpfr.so or libmpc.so
>
>   Isn't the dynamic linker, ld.so and ld-linux.so2 the same thing?

Yes.

But ld is not the same thing as ld.so and my point is that ld might
have been able to find the libraries, but ld.so could not.

But this still seems like a waste of time until you know what the
actual error was.  I still suggest looking in
$target/libgcc/config.log to see if it's in there.

Re: gcc-4.6.0 breaks

[Bill Cunningham]

Jonathan Wakely wrote:

>> No. Although I didn't manually use the switches with gcc to find gmp
>> mpc 
>> or mpfc. The configure script seemed to me atleast to be able to
>> find those libraries I put in /usr/local. No when compiling mpc I
>> believe it was it needed to know where gmp was. The script needed
>> help there. gmp mpc and mpfc were the only things gcc to my
>> knowledge needed to compile and run. 
> 
> There's a difference between configure being able to find the
> libraries to link an executable and the dynamic linker being able to
> find those libraris to *run* an executable.
> 
> If /usr/local/lib is not in your ldconfig cache then it will not find
> libgmp.so etc. so you will need to reconfigure ldconfig, or set
> LD_LIBRARY_PATH.
> 
> Of course since you don't know what the actual error was this is all
> just wild speculation.  You need to know what the error was and if it
> told you to look in config.log, which would have more information.
> 
> Try looking in the build directory for $target/libgcc/config.log which
> I suspect will show that ld.so couldn't find one of libgmp.so,
> libmpfr.so or libmpc.so

    Isn't the dynamic linker, ld.so and ld-linux.so2 the same thing?

Bill

Re: gcc-4.6.0 breaks

[Jonathan Wakely]

On 2 June 2011 03:25, Bill Cunningham wrote:
> Dennis Clarke wrote:
>>>
>>>    My gcc-4.6.0 compilation after compiling for quite some time
>>> decided to
>>> break with an error to stderr of something like not recognized object
>>> files.
>>> I am compiling with gcc-3.3.6. This can sure be nerve racking. Does
>>> anyone know what this means?
>>>
>>
>> How would anyone even begin to help with this email content ?
>>
>> Okay, so let's see, so you're running Debian Linux on a DEC alpha and
>> you are trying to bootstrap GCC 4.6.0 while using GCC 3.3.6 with
>> binutils maybe maybe not and your configure line was black magic
>> while facing east with a dead cat under a rising moon ?
>>
>> My guess is that you need to face west.
>>
>> Try again with some content please !
>
>   I didn't get a log of stderr. I wish I had it turned on now. I'm using an
> old RH 9 for the i686 and binutils-2.21 along with gcc-3.3.6. I have have
> gmp mpfc and mpc all installed and I am using make-3.82. I was rather
> surprised when it broke. If I try again I will get a stderr log.

Couldn't you have used a pencil to write down the actual text of the error?


Was it this error: http://gcc.gnu.org/wiki/FAQ#configure_suffix ?
Did you read the installation docs about --with-gmp and ensuring the
dynamic linker can find libgmp etc?

See http://advogato.org/person/redi/diary/240.html and/or
http://advogato.org/person/redi/diary/229.html

Re: gcc-4.6.0 breaks

[Bill Cunningham]

Dennis Clarke wrote:
>>     My gcc-4.6.0 compilation after compiling for quite some time
>> decided to
>> break with an error to stderr of something like not recognized object
>> files.
>> I am compiling with gcc-3.3.6. This can sure be nerve racking. Does
>> anyone know what this means?
>>
>
> How would anyone even begin to help with this email content ?
>
> Okay, so let's see, so you're running Debian Linux on a DEC alpha and
> you are trying to bootstrap GCC 4.6.0 while using GCC 3.3.6 with
> binutils maybe maybe not and your configure line was black magic
> while facing east with a dead cat under a rising moon ?
>
> My guess is that you need to face west.
>
> Try again with some content please !

    I didn't get a log of stderr. I wish I had it turned on now. I'm using 
an old RH 9 for the i686 and binutils-2.21 along with gcc-3.3.6. I have have 
gmp mpfc and mpc all installed and I am using make-3.82. I was rather 
surprised when it broke. If I try again I will get a stderr log.

Bill

Re: gcc-4.6.0 breaks

[Dennis Clarke]

>     My gcc-4.6.0 compilation after compiling for quite some time decided
> to
> break with an error to stderr of something like not recognized object
> files.
> I am compiling with gcc-3.3.6. This can sure be nerve racking. Does anyone
> know what this means?
>

How would anyone even begin to help with this email content ?

Okay, so let's see, so you're running Debian Linux on a DEC alpha and you
are trying to bootstrap GCC 4.6.0 while using GCC 3.3.6 with binutils
maybe maybe not and your configure line was black magic while facing east
with a dead cat under a rising moon ?

My guess is that you need to face west.

Try again with some content please !

Dennis

gcc-4.6.0 breaks

[Bill Cunningham]

    My gcc-4.6.0 compilation after compiling for quite some time decided to 
break with an error to stderr of something like not recognized object files. 
I am compiling with gcc-3.3.6. This can sure be nerve racking. Does anyone 
know what this means?

Bill