Install error on Suse 11.1

[henrik22]

Hi,
on Suse 11.1 with gcc 4.3.2 installation stops early in pari-2.3.3.p0
with

/bin/sh: symbol lookup error: /home/henrik/Pakete/sage-3.2.1/local/
lib/libreadline.so.5: undefined symbol: PC

I am running on an x64_64 dual core notebook with Suse 11.1rc1.
Any help appreciated, thank you very much in advance! Henrik


>>> Last lines of install.log:
...
pari-2.3.3.p0/SPKG.txt
pari-2.3.3.p0/.hgignore
Finished extraction
****************************************************
Host system
uname -a:
Linux linux-wccv 2.6.27.7-8-default #1 SMP 2008-11-25 00:02:37 +0100
x86_64 x86_64 x86_64 GNU/Linux
****************************************************
****************************************************
GCC Version
gcc -v
Using built-in specs.
Target: x86_64-suse-linux
Configured with: ../configure --prefix=/usr --infodir=/usr/share/info
--mandir=/usr/share/man --libdir=/usr/lib64 --libexecdir=/usr/lib64 --
enable-languages=c,c++,objc,fortran,obj-c++,java,ada --enable-
checking=release --with-gxx-include-dir=/usr/include/c++/4.3 --enable-
ssp --disable-libssp --with-bugurl=http://bugs.opensuse.org/ --with-
pkgversion='SUSE Linux' --disable-libgcj --disable-libmudflap --with-
slibdir=/lib64 --with-system-zlib --enable-__cxa_atexit --enable-
libstdcxx-allocator=new --disable-libstdcxx-pch --enable-version-
specific-runtime-libs --program-suffix=-4.3 --enable-linux-futex --
without-system-libunwind --with-cpu=generic --build=x86_64-suse-linux
Thread model: posix
gcc version 4.3.2 [gcc-4_3-branch revision 141291] (SUSE Linux)
****************************************************
/bin/sh: symbol lookup error: /home/henrik/Pakete/sage-3.2.1/local/lib/
libreadline.so.5: undefined symbol: PC

real 0m0.001s
user 0m0.000s
sys 0m0.000s
sage: An error occurred while installing pari-2.3.3.p0
Please email sage-devel http://groups.google.com/group/sage-devel

[mabshoff]

henrik22 wrote:
> Hi,

Hi Henrik,

> on Suse 11.1 with gcc 4.3.2 installation stops early in pari-2.3.3.p0
> with
>
> /bin/sh: symbol lookup error: /home/henrik/Pakete/sage-3.2.1/local/
> lib/libreadline.so.5: undefined symbol: PC
>
> I am running on an x64_64 dual core notebook with Suse 11.1rc1.
> Any help appreciated, thank you very much in advance! Henrik

Please post a link to install.log - do not attach it to the email. We
cannot debug this without the install.log. My guess would be that pari
picks up a global readline.h.

Cheers,

Michael

[henrik22]

Thank you for the hint. In fact, compilation continues if I
remove the just created readline from sage-3.2.2/local/lib, so that
the pari compilation
is forced to use the version 5.2 already on my computer (/lib64/
libreadline.so.5.2,
including header files in /usr/include/readline).

Where may I post the install.log file to? File upload here in this
group does not work (for me?).
Regards, Henrik

P.S. The problem is identical for sage 3.2.1 and 3.2.2.

[mabshoff]

On Dec 19, 12:56 pm, henrik22 <heb...@arcor.de> wrote:

Hi,

> Thank you for the hint. In fact, compilation continues if I
> remove the just created readline from sage-3.2.2/local/lib, so that
> the pari compilation
> is forced to use the version 5.2 already on my computer (/lib64/
> libreadline.so.5.2,
> including header files in /usr/include/readline).

Well, that is probably not the real fix and not a good idea, but we
will see.

> Where may I post the install.log file to? File  upload here in this
> group does not work (for me?).

Yes, I didn't mean attaching the files to the Google group which
requires admin privileges. Any place on the web is fine, but make sure
to compress the old install.log when the build failed. As a last
resort you can send it to me compressed off list, but that is truly
the last resort.

> Regards, Henrik
>
> P.S. The problem is identical for sage 3.2.1 and 3.2.2.

Ok, I am downloading OpenSUSE 11.1 as a VMWare image, so hopefully I
can fix it locally. That should be much easier to fix that way than by
remote diagnostics.

Cheers,

Michael

[mabshoff]

Hi,

no need to send me any logs since I can reproduce this now. The
problem is that bash as well as sh on OpenSUSE are dynamically linked
against readline, so one we source local/bin/sage-env from an spkg
during build and we do have a compiled and installed libreadline.so
things blow up with a missing symbol. I am poking around what is wrong
here, but there is no obvious fix.

Cheers,

Michael

[fhivert]

Dear Michael,

> I am poking around what is wrong
> here, but there is no obvious fix.

I just ran into the same problem. If I understand correctly,
The suggested fix of removing the local copy of readline
will compile but result in a non working sage, right ?

By the way if it if any use I'm running a x64_64 dual core notebook
with the new Suse 11.1

Cheers,

Florent

[mabshoff]

On Dec 20, 7:56 am, fhivert <florent.hiv...@univ-rouen.fr> wrote:
>    Dear Michael,

Hi Florent,

> > I am poking around what is wrong
> > here, but there is no obvious fix.
>
> I just ran into the same problem. If I understand correctly,
> The suggested fix of removing the local copy of readline
> will compile but result in a non working sage, right ?

It will probably work [I killed my build last night], but the
resulting Sage could behave oddly since the readline from the system
is slightly different than the one we compile. I have seen odd
segfaults with tab completion where in the end we agreed that it was
because for example Python used to be compiled against the system
readline. That has been fixed.

> By the way if it if any use I'm running a x64_64 dual core  notebook
> with the new Suse 11.1

Ok.

> Cheers,
>
> Florent

I have thought about how to fix this and there are various solutions,
all of which suck one way or the other. So once I come up with some
better idea I will let you know :)

Cheers,

Michael

[fhivert]

Dear Michael,

> It will probably work [I killed my build last night], but the
> resulting Sage could behave oddly since the readline from the system
> is slightly different than the one we compile.

So it's worth giving a try rather than being stuck with no sage at
all. By the way I had the same mistake when trying to install a
precompiled version for openSuSE 10.3

> I have thought about how to fix this and there are various solutions,
> all of which suck one way or the other. So once I come up with some
> better idea I will let you know :)

I'll keep listening,

Cheers,

Florent

[henrik22]

Dear all,
at least it seems to work for me now after
1) removing the sage-compiled readline
2) replacing the systems iconv by a new one (configure of R
complained about that)
there still seem to be problems with iconv, but probably that is
another, independent thread...

Best regards and thanks for such a huge bundle of software (whish i
could meaningfully use 10% of it),
Henrik

[mabshoff]

On Dec 20, 11:03 am, henrik22 <heb...@arcor.de> wrote:
> Dear all,

Hi Henrik,

> at least it seems to work for me now after
>  1) removing the sage-compiled readline

I got an updated spkg that doesn't build readline on systems where
bash is linked dynamically against readline. That seems to be the
least invasive fix for now.

>  2) replacing the systems iconv by a new one (configure of R
> complained about that)

I am not sure what you mean here since the iconv in OpenSUSE 11.1
should certainly be current enough, especially considering that we
ship a somewhat older R release.

> there still seem to be problems with iconv, but probably that is
> another, independent thread...

I am tempted to disable iconv support in R for now, but I will see
what R complains about. My feeling is that R's iconv detection is bad/
broken, but we will see.

> Best regards and thanks for such a huge bundle of software (whish i
> could meaningfully use 10% of it),

:)

> Henrik

Cheers,

Michael

[mabshoff]

Ok, I got Sage 3.2.2 to build and pass all doctests on a 32 bit
OpenSUSE 11.1, but there is no good fix yet to resolve the build issue
automatically. One problem is that if we use the system readline
Singular as well as clisp fail (I did not see any issue with R and
iconv) and if we delete libreadline.so Python's readline extension
also isn't build properly, so unless we force the install of readline-
devel my current workaround fix won't work since the readline headers
are missing. We can copy the system readline into $SAGE_LOCAL/lib as a
last resort, but that is potentially risky. I will keep digging to see
if I can fix this in a more elegant way.

Cheers,

Michael