Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Segher Boessenkool via cfarm-users]

On Tue, Feb 20, 2018 at 07:21:42AM -0500, Jeffrey Walton via cfarm-users wrote:
> On Tue, Feb 20, 2018 at 6:54 AM, Jonathan Wakely  
> wrote:
> Now I am even more confused.
> 
> The admins install stiff in /opt/cfarm. I used to look in /usr/local
> for the upgrades, but I was told to look in /opt/cfarm instead.
> 
> How is that randomized shit that nobody knows anything about?

The *-latest things are supposed to be the latest releases of programs
we care about (but "latest" of course means "latest at the time it was
installed" ;-) ).  Anything else is, well, something else.


Segher
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Segher Boessenkool via cfarm-users]

On Tue, Feb 20, 2018 at 11:54:34AM +, Jonathan Wakely via cfarm-users wrote:
> On 19 February 2018 at 22:15, Jeffrey Walton via cfarm-users wrote:
> >>> I used a GNUmakefile:
> >>>
> >>> CXX=/opt/cfarm/gcc7rc2/bin/g++ make -j 20
> >>>
> >>> Nothing special, as far as I know. Just using the latest compiler.
> 
> That's not the latest compiler. It's not even an officially released
> compiler, and it doesn't even seem to have been built for the right OS
> (it predates the upgrade to CentOS).

No, it is more than two months younger.  But it was built on a different
machine (and different OS).  As I explained a few times already, I put
it there to help people who cannot easily build a compiler with Ada
support (because centos does not ship that currently).

And it in fact works fine for that.  I used it yesterday to build that
gcc8-r257824, which will work for more things (it has a working libstdc++,
or at least the GCC testsuite didn't complain).

GCC 7 RC2 is essentially identical to 7.0.1 fwiw.


Segher
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Jonathan Wakely via cfarm-users]

On 20 February 2018 at 12:21, Jeffrey Walton  wrote:
> On Tue, Feb 20, 2018 at 6:54 AM, Jonathan Wakely  
> wrote:
>> ...
 That is special.  /opt/cfarm is not installed as an alternate system
 compiler.  It's not even the latest compiler.  If you choose to
 override the defaults and refer to a random compiler whose origin you
 don't know, you may encounter unintended consequences, as you have
 discovered.
>>>
>>> Oh, my bad. I thought the latest stuff was installed in /opt/cfarm.
>>>
>>> So I am clear: we are _not_ supposed to use stuff in /opt/cfarm. Is
>>> that correct?
>>
>> That's not what he said, he said "If you choose to override the
>> defaults and refer to a random compiler whose origin you don't know,
>> you may encounter unintended consequences".
>
> Now I am even more confused.
>
> The admins install stiff in /opt/cfarm. I used to look in /usr/local
> for the upgrades, but I was told to look in /opt/cfarm instead.
>
> How is that randomized shit that nobody knows anything about?

Nobody said it's randomized shit that nobody knows about, it's a
compiler YOU don't know anything about. The suggestion was that if
YOU, Jeff, do not know what /opt/cfarm/gcc7rc2 is or where it comes
from, maybe you shouldn't be relying on it, and if it doesn't work,
stop using it (as I've already said, it doesn't work because it was
built for a different OS). Unless you specifically need to reproduce a
bug with GCC 7.0.1 RC2 then that version wouldn't be very useful even
if it was built for the right OS.

The other packages in /opt/cfarm should be OK to use if you need GMP, MPFR etc.
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Jeffrey Walton via cfarm-users]

On Tue, Feb 20, 2018 at 6:54 AM, Jonathan Wakely  wrote:
> ...
>>> That is special.  /opt/cfarm is not installed as an alternate system
>>> compiler.  It's not even the latest compiler.  If you choose to
>>> override the defaults and refer to a random compiler whose origin you
>>> don't know, you may encounter unintended consequences, as you have
>>> discovered.
>>
>> Oh, my bad. I thought the latest stuff was installed in /opt/cfarm.
>>
>> So I am clear: we are _not_ supposed to use stuff in /opt/cfarm. Is
>> that correct?
>
> That's not what he said, he said "If you choose to override the
> defaults and refer to a random compiler whose origin you don't know,
> you may encounter unintended consequences".

Now I am even more confused.

The admins install stiff in /opt/cfarm. I used to look in /usr/local
for the upgrades, but I was told to look in /opt/cfarm instead.

How is that randomized shit that nobody knows anything about?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Jonathan Wakely via cfarm-users]

On 19 February 2018 at 22:15, Jeffrey Walton via cfarm-users wrote:
>>> I used a GNUmakefile:
>>>
>>> CXX=/opt/cfarm/gcc7rc2/bin/g++ make -j 20
>>>
>>> Nothing special, as far as I know. Just using the latest compiler.

That's not the latest compiler. It's not even an officially released
compiler, and it doesn't even seem to have been built for the right OS
(it predates the upgrade to CentOS).

> I'm trying to reproduce a problem under Fedora 28.

Then using GCC 7.0.1 RC2 on CentOS 7 doesn't make much sense. It's a
different GCC version and a different glibc version.

Using /opt/cfarm/gcc8-r257824/bin/g++ would be slightly better (which
it seems you have also tried).

> The problem is not present under the default compiler, which is GCC 4.8.5.



>> That is special.  /opt/cfarm is not installed as an alternate system
>> compiler.  It's not even the latest compiler.  If you choose to
>> override the defaults and refer to a random compiler whose origin you
>> don't know, you may encounter unintended consequences, as you have
>> discovered.
>
> Oh, my bad. I thought the latest stuff was installed in /opt/cfarm.
>
> So I am clear: we are _not_ supposed to use stuff in /opt/cfarm. Is
> that correct?

That's not what he said, he said "If you choose to override the
defaults and refer to a random compiler whose origin you don't know,
you may encounter unintended consequences".
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Jeffrey Walton via cfarm-users]

>>> On Mon, Feb 19, 2018 at 02:49:47PM -0500, Jeffrey Walton via cfarm-users 
>>> wrote:
 /opt/cfarm/gcc7rc2/bin/g++ -o cryptest.exe -DNDEBUG -g2 -O3 -fPIC
 -pthread -pipe adhoc.o test.o bench1.o bench2.o validat0.o validat1.o
 validat2.o validat3.o validat4.o datatest.o regtest1.o regtest2.o
 regtest3.o dlltest.o fipsalgt.o ./libcryptopp.a
 /home/guerby/opt/cfarm/gcc7rc2/bin/../lib/gcc/powerpc64le-unknown-linux-gnu/7.0.1/../../../../lib64/libstdc++.so:
 undefined reference to `__cxa_thread_atexit_impl@GLIBC_2.18'
 collect2: error: ld returned 1 exit status
 make: *** [cryptest.exe] Error 1
>>>
>>> How/why do you get anything from /home/guerby?  /opt/cfarm/gcc7rc2 was
>>> build on a different machine (and it is only there so people can use it
>>> as an easy starting point for building ada: the system compiler on 112
>>> does not support that) so there is no way it refers to there.
>>
>> Hmm, I'm not sure. I'm on GCC112, and it is in /opt/cfarm:
>>
>> $ ls /opt/cfarm/
>> autoconf-latestftp krb5-latest  mpfr-latest python3-latest
>> bison-latest   gcc7rc2 m4-latestopenssl-latest  release
>> cyrus-sasl-latest  gcc-latest  make-latest  perl5-latestzlib-latest
>> db-latest  gmp-latest  mpc-latest   python2-latest
>>
>>> I know nothing about the build system you used, but it seems buggy.
>>
>> I used a GNUmakefile:
>>
>> CXX=/opt/cfarm/gcc7rc2/bin/g++ make -j 20
>>
>> Nothing special, as far as I know. Just using the latest compiler.
>
> Why are you using /opt/cfarm instead of the default system compiler?

I'm trying to reproduce a problem under Fedora 28.

The problem is not present under the default compiler, which is GCC 4.8.5.

> That is special.  /opt/cfarm is not installed as an alternate system
> compiler.  It's not even the latest compiler.  If you choose to
> override the defaults and refer to a random compiler whose origin you
> don't know, you may encounter unintended consequences, as you have
> discovered.

Oh, my bad. I thought the latest stuff was installed in /opt/cfarm.

So I am clear: we are _not_ supposed to use stuff in /opt/cfarm. Is
that correct?

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[David Edelsohn via cfarm-users]

On Mon, Feb 19, 2018 at 4:20 PM, Jeffrey Walton via cfarm-users
 wrote:
> On Mon, Feb 19, 2018 at 4:12 PM, Segher Boessenkool
>  wrote:
>> Hi,
>>
>> On Mon, Feb 19, 2018 at 02:49:47PM -0500, Jeffrey Walton via cfarm-users 
>> wrote:
>>> /opt/cfarm/gcc7rc2/bin/g++ -o cryptest.exe -DNDEBUG -g2 -O3 -fPIC
>>> -pthread -pipe adhoc.o test.o bench1.o bench2.o validat0.o validat1.o
>>> validat2.o validat3.o validat4.o datatest.o regtest1.o regtest2.o
>>> regtest3.o dlltest.o fipsalgt.o ./libcryptopp.a
>>> /home/guerby/opt/cfarm/gcc7rc2/bin/../lib/gcc/powerpc64le-unknown-linux-gnu/7.0.1/../../../../lib64/libstdc++.so:
>>> undefined reference to `__cxa_thread_atexit_impl@GLIBC_2.18'
>>> collect2: error: ld returned 1 exit status
>>> make: *** [cryptest.exe] Error 1
>>
>> How/why do you get anything from /home/guerby?  /opt/cfarm/gcc7rc2 was
>> build on a different machine (and it is only there so people can use it
>> as an easy starting point for building ada: the system compiler on 112
>> does not support that) so there is no way it refers to there.
>
> Hmm, I'm not sure. I'm on GCC112, and it is in /opt/cfarm:
>
> $ ls /opt/cfarm/
> autoconf-latestftp krb5-latest  mpfr-latest python3-latest
> bison-latest   gcc7rc2 m4-latestopenssl-latest  release
> cyrus-sasl-latest  gcc-latest  make-latest  perl5-latestzlib-latest
> db-latest  gmp-latest  mpc-latest   python2-latest
>
>> I know nothing about the build system you used, but it seems buggy.
>
> I used a GNUmakefile:
>
> CXX=/opt/cfarm/gcc7rc2/bin/g++ make -j 20
>
> Nothing special, as far as I know. Just using the latest compiler.

Why are you using /opt/cfarm instead of the default system compiler?

That is special.  /opt/cfarm is not installed as an alternate system
compiler.  It's not even the latest compiler.  If you choose to
override the defaults and refer to a random compiler whose origin you
don't know, you may encounter unintended consequences, as you have
discovered.

Thanks, David
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Jeffrey Walton via cfarm-users]

On Mon, Feb 19, 2018 at 4:12 PM, Segher Boessenkool
 wrote:
> Hi,
>
> On Mon, Feb 19, 2018 at 02:49:47PM -0500, Jeffrey Walton via cfarm-users 
> wrote:
>> /opt/cfarm/gcc7rc2/bin/g++ -o cryptest.exe -DNDEBUG -g2 -O3 -fPIC
>> -pthread -pipe adhoc.o test.o bench1.o bench2.o validat0.o validat1.o
>> validat2.o validat3.o validat4.o datatest.o regtest1.o regtest2.o
>> regtest3.o dlltest.o fipsalgt.o ./libcryptopp.a
>> /home/guerby/opt/cfarm/gcc7rc2/bin/../lib/gcc/powerpc64le-unknown-linux-gnu/7.0.1/../../../../lib64/libstdc++.so:
>> undefined reference to `__cxa_thread_atexit_impl@GLIBC_2.18'
>> collect2: error: ld returned 1 exit status
>> make: *** [cryptest.exe] Error 1
>
> How/why do you get anything from /home/guerby?  /opt/cfarm/gcc7rc2 was
> build on a different machine (and it is only there so people can use it
> as an easy starting point for building ada: the system compiler on 112
> does not support that) so there is no way it refers to there.

Hmm, I'm not sure. I'm on GCC112, and it is in /opt/cfarm:

$ ls /opt/cfarm/
autoconf-latestftp krb5-latest  mpfr-latest python3-latest
bison-latest   gcc7rc2 m4-latestopenssl-latest  release
cyrus-sasl-latest  gcc-latest  make-latest  perl5-latestzlib-latest
db-latest  gmp-latest  mpc-latest   python2-latest

> I know nothing about the build system you used, but it seems buggy.

I used a GNUmakefile:

CXX=/opt/cfarm/gcc7rc2/bin/g++ make -j 20

Nothing special, as far as I know. Just using the latest compiler.

Jeff
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Jonathan Wakely via cfarm-users]

On 19 February 2018 at 20:17, Andreas Schwab via cfarm-users
 wrote:
> On Feb 19 2018, Jeffrey Walton via cfarm-users 
>  wrote:
>
>> Compile appears OK but link fails:
>>
>> /opt/cfarm/gcc7rc2/bin/g++ -o cryptest.exe -DNDEBUG -g2 -O3 -fPIC
>> -pthread -pipe adhoc.o test.o bench1.o bench2.o validat0.o validat1.o
>> validat2.o validat3.o validat4.o datatest.o regtest1.o regtest2.o
>> regtest3.o dlltest.o fipsalgt.o ./libcryptopp.a
>> /home/guerby/opt/cfarm/gcc7rc2/bin/../lib/gcc/powerpc64le-unknown-linux-gnu/7.0.1/../../../../lib64/libstdc++.so:
>> undefined reference to `__cxa_thread_atexit_impl@GLIBC_2.18'
>
> gcc112 has glibc 2.17 installed, but this compiler was built against
> glibc 2.18 or newer.  That cannot work.

Presumably that copy of GCC was built when the machine was running
Fedora, before it was reinstalled as CentOS. You need to use a GCC
built for the current OS.
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Jonathan Wakely via cfarm-users]

On 19 February 2018 at 19:49, Jeffrey Walton via cfarm-users
 wrote:
> Hi Everyone,
>
> We are trying to reproduce an issue reported on Fedora 28 ppc64-le. I
> don't know the version of GCC on Fedora 28 at the moment.

It's currently a recent GCC 8.0.1 snapshot.
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users

Re: [cfarm-users] gcc112/ppc64-le, and undefined reference to __cxa_thread_atexit_impl

[Andreas Schwab via cfarm-users]

On Feb 19 2018, Jeffrey Walton via cfarm-users 
 wrote:

> Compile appears OK but link fails:
>
> /opt/cfarm/gcc7rc2/bin/g++ -o cryptest.exe -DNDEBUG -g2 -O3 -fPIC
> -pthread -pipe adhoc.o test.o bench1.o bench2.o validat0.o validat1.o
> validat2.o validat3.o validat4.o datatest.o regtest1.o regtest2.o
> regtest3.o dlltest.o fipsalgt.o ./libcryptopp.a
> /home/guerby/opt/cfarm/gcc7rc2/bin/../lib/gcc/powerpc64le-unknown-linux-gnu/7.0.1/../../../../lib64/libstdc++.so:
> undefined reference to `__cxa_thread_atexit_impl@GLIBC_2.18'

gcc112 has glibc 2.17 installed, but this compiler was built against
glibc 2.18 or newer.  That cannot work.

Andreas.

-- 
Andreas Schwab, sch...@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."
___
cfarm-users mailing list
cfarm-users@lists.tetaneutral.net
https://lists.tetaneutral.net/listinfo/cfarm-users