Building of sage 9.2 stopped because pynac doesn't see libflint

[Tobias]

Hallo Everybody,

on OpenSuse Tumbleweed I try building sage 9.2 without most of packages from
system. The make command stopped compiling pynac because "the programm needs
libflint", but flint was already compiled by make command.

Then I wanted to start make command again with pynac, but make skipped pynac
and started with the next package.

What should I do? I would probably restart building and configure with --with-
system-flint=yes and --with-system-flintqs=yes
But it took so long to come here ...

What would you say?

Kind regards,
Tobias

from log file
> sage-9.2/local/var/tmp/sage/build/pynac-0.7.26.sage-2020-04-03.p0/src/config
>.log

> configure:17041: gcc -o conftest -g -O2
> -Wl,-rpath-link,/home/tobias/bin/sage-9.2/local/lib
> -L/home/tobias/bin/sage-9.2/local/lib
> -Wl,-rpath,/home/tobias/bin/sage-9.2/local/lib conftest.c -lflint -lgmp
> >&5
> /usr/lib64/gcc/x86_64-suse-linux/10/../../../../x86_64-suse-linux/bin/ld:
> warning: libgmp.so.23, needed by
> /home/tobias/bin/sage-9.2/local/lib/libflint.so, may conflict with
> libgmp.so.10
> /usr/lib64/gcc/x86_64-suse-linux/10/../../../../x86_64-suse-linux/bin/ld:
> _ZN3NTL9zz_pEInfoE: TLS definition in /usr/lib64/libntl.so.43 section .tbss
> mismatches non-TLS reference in
> /home/tobias/bin/sage-9.2/local/lib/libflint.so
> /usr/lib64/gcc/x86_64-suse-linux/10/../../../../x86_64-suse-linux/bin/ld:
> /usr/lib64/libntl.so.43: error adding symbols: bad value collect2: error:
> ld returned 1 exit status
> configure:17041: $? = 1
>
> configure: failed program was:
> | /* confdefs.h */
> | #define PACKAGE_NAME "pynac"
> | #define PACKAGE_TARNAME "pynac"
> | #define PACKAGE_VERSION "0.7.26.sage-2020-04-03"
> | #define PACKAGE_STRING "pynac 0.7.26.sage-2020-04-03"
> | #define PACKAGE_BUGREPORT "<pynac...@googlegroups.com>"
> | #define PACKAGE_URL ""
> | #define PACKAGE "pynac"
> | #define VERSION "0.7.26.sage-2020-04-03"
> | #define ARCHIVE_VERSION 3
> | #define ARCHIVE_AGE 0
> | #define STDC_HEADERS 1
> | #define HAVE_SYS_TYPES_H 1
> | #define HAVE_SYS_STAT_H 1
> | #define HAVE_STDLIB_H 1
> | #define HAVE_STRING_H 1
> | #define HAVE_MEMORY_H 1
> | #define HAVE_STRINGS_H 1
> | #define HAVE_INTTYPES_H 1
> | #define HAVE_STDINT_H 1
> | #define HAVE_UNISTD_H 1
> | #define HAVE_DLFCN_H 1
> | #define LT_OBJDIR ".libs/"
> | #define HAVE_GMP_H 1
> | #define HAVE_FLINT_FMPQ_POLY_H 1
> | /* end confdefs.h. */
> |
> | /* Override any GCC internal prototype to avoid an error.
> |
> | Use char because int might match the return type of a GCC
> | builtin and then its argument prototype would still apply. */
> |
> | #ifdef __cplusplus
> | extern "C"
> | #endif
> | char fmpq_get_mpz_frac ();
> | int
> | main ()
> | {
> | return fmpq_get_mpz_frac ();
> |
> | ;
> | return 0;
> |
> | }
>
> configure:17058: result: no
> configure:17065: error: This package needs libflint

or from file
> sage-9.2/logs/pkgs/pynac-0.7.26.sage-2020-04-03.p0.log

> checking for library containing __gmpz_get_str... -lgmp
> checking flint/fmpq_poly.h usability... yes
> checking flint/fmpq_poly.h presence... yes
> checking for flint/fmpq_poly.h... yes
> checking for library containing fmpq_get_mpz_frac... no
> configure: error: This package needs libflint
> ****************************************************************************
> ****************************************************************************
> ************************************************************************
> Error configuring pynac-0.7.26.sage-2020-04-03.p0
> See the file
>
> /home/tobias/bin/sage-9.2/local/var/tmp/sage/build/pynac-0.7.26.sage-202
> 0-04-03.p0/src/config.log>
> for details.
> ****************************************************************************
> ****************************************************************************
> ************************************************************************

[Dima Pasechnik]

On Fri, 27 Nov 2020, 14:21 Tobias, <tobwe...@web.de> wrote:

Hallo Everybody,

on OpenSuse Tumbleweed I try building sage 9.2 without most of packages from
system. The make command stopped compiling pynac because "the programm needs
libflint", but flint was already compiled by make command.



Then I wanted to start make command again with pynac, but make skipped pynac
and started with the next package.


What should I do? I would probably restart building and configure with --with-
system-flint=yes and --with-system-flintqs=yes
But it took so long to come here ...


What would you say?

I guess you have flint installed systemwide, as well, and pynac gets confused as to which flint to use.

--
You received this message because you are subscribed to the Google Groups "sage-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sage-support...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/2410372.RVVYDCxLDo%40localhost.localdomain.

[Tobias Weiss]

OK! I just uninstalled flint and did configure without system packages again (after make clean). I can tell you if it works tomorrow or later, because of compile time ...

Well, I first tried to compile with system packages, but this made an error, when making sagelib - beacause of pari, I think.

Thank you!  I hope this works,
Tobias

[Dima Pasechnik]

On Fri, Nov 27, 2020 at 5:44 PM Tobias Weiss <tobwe...@web.de> wrote:
>
> OK! I just uninstalled flint and did configure without system packages again (after make clean). I can tell you if it works tomorrow or later, because of compile time ...
>
> Well, I first tried to compile with system packages, but this made an error, when making sagelib - beacause of pari, I think.

We very much welcome reports of failures like this, as we strive to
unvendor as much as possible.
Skipping all the system packages is surely a huge overkill - e.g. why
on Earth would you need to build gcc
on a regular Linux system?

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/23ae3a10-bf65-424e-9051-b0e89fa1d1d6n%40googlegroups.com.

[Tobias Weiss]

I didn't want to find out, what is the problem with sagelib. The error was from the system packages. So I wanted no conflicts between system wide packages and the packages from the tarball. So I configured like this:

./configure --with-system-arb=no --with-system-boost=no --with-system-brial=no --with-system-cddlib=no --with-system-cliquer=no --with-system-ecm=no --with-system-eclib=no --with-system-fflas_ffpack=no --
with-system-flint=no --with-system-flintqs=no --with-system-fplll=no --with-system-freetype=no --with-system-gc=no --with-system-giac=no --with-system-givaro=no --with-system-glpk=no --with-system-gmp=no --with-system-gsl=no
--with-system-iconv=no --with-system-iml=no --with-system-isl=no --with-system-lcalc=no --with-system-libbraiding=no --with-system-libatomic_ops=no --with-system-libffi=no --with-system-libgd=no --with-system-libsemigroups=
no --with-system-lrcalc=no --with-system-m4rie=no --with-system-m4ri=no --with-system-mpc=no --with-system-mpfi=no --with-system-mpfr=no --with-system-mpir=no --with-system-ninja_build=no --with-system-ntl=no --with-system-o
penblas=no --with-system-pandoc=no --with-system-pari=no --with-system-pari_galdata=no --with-system-pari_seadata_small=no --with-system-pcre=no --with-system-perl_term_readline_gnu=no --with-system-planarity=no --with-syste
m-ppl=no --with-system-rw=no --with-system-readline=no --with-system-suitesparse=no --with-system-symmetrica=no --with-system-yasm=no --with-system-zeromq=no --with-system-zn_poly=no --with-system-pynac=no

Sure, I wanted to use gcc from system. But it was also compiled newly.

Shall I reproduce the error with sagelib, when I have time?

Tobias

[Tobias Weiss]

... how does compiling work? Needs the compiler also the binary of some library or only the library-devel package with the code? For most packages I only installed the devel package ...

Tobias

[Tobias Weiss]

... I thought there must be a reason I can install the development package without binary package. But no it makes no sense to me. Can this be the reason of both complile errors?

Tobias

[Dima Pasechnik]

On Fri, Nov 27, 2020 at 7:21 PM Tobias Weiss <tobwe...@web.de> wrote:
>
> ... I thought there must be a reason I can install the development package without binary package. But no it makes no sense to me. Can this be the reason of both complile errors?
>
> Tobias
>
> Tobias Weiss schrieb am Freitag, 27. November 2020 um 20:15:24 UTC+1:
>>
>> ... how does compiling work? Needs the compiler also the binary of some library or only the library-devel package with the code? For most packages I only installed the devel package ...

I don't know much about OpenSUSE, but some distributions typically
have <library>-dev packages dependent on <library> packages, and
some (e.g. Fedora) do not, i.e. you always need to install both
<library>-dev (or <library>-devel) and <library>.
It seems to me that OpenSUSE is an rpm-based distro, thus it would be
like Fedora, and require <library>-devel and <library> in some cases.


If you just run ./configure
it will print at the end some messages about packages that are
recommended to be installed (after guessing, hopefully correctly
the kind of Linux system it is on)

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/773dfae9-620c-456e-9953-7a94e5270c6bn%40googlegroups.com.

[Tobias Weiss]

The Problem is solved now: However, make clean didn't delete the whole make history. And even now pynac was build. So, uninstalling the system wide flint-devel package is the solution!!!
There is still the question with the binary files - but this I also can find out somewhere else ...

Thanks,

Tobias

[Tobias Weiss]

... it is a bit strange: now pynac is build before sagelib on the beginning. Is pynac build two times?

[Tobias Weiss]

OK, so I stop the current build. Then I install for every dependency both - devel and binary package - and retry the build tomorrow.

Then it don't makes sence, that the binary is not installed automatically when I install the source package - but this is an Opensuse problem. Once, it was like this ...

Thank you!

Tobias

[Tobias Weiss]

I still have a question: Keeping in mind Tumbleweed is a rolling release of Opensuse - is it clever to build a fixed binary? Will this binary persist with the rolling software updates?

Maybe that's why the dependencies of source and binary package don't work? I will test these dependencies tomorrow at all, so that I don't tell you wrong things and the dependencies work at all ...

Tobias

[Dima Pasechnik]

On Fri, Nov 27, 2020 at 9:36 PM Tobias Weiss <tobwe...@web.de> wrote:
>
> I still have a question: Keeping in mind Tumbleweed is a rolling release of Opensuse - is it clever to build a fixed binary? Will this binary persist with the rolling software updates?

we regularly test SageMath on Gentoo, which is a rolling release system, too.
Needless to say, once in a while one might need to rebuild something,
if a library SageMath depends on was updated
in an incompatible way.

>
> Maybe that's why the dependencies of source and binary package don't work?

no, it's not they don't work, it's a design choice, leading to fewer updates;
often only a binary package changes, then if it is a dynamic library
then nothing needs to be rebuilt.
(perhaps - I am not sure here; perhaps it's just a historical choice
that stayed)

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/76e63d8e-6530-4c2d-9321-1fb8a1adb01cn%40googlegroups.com.

[Tobias Weiss]

Is it possible to create a rpm file after build? If I install sage with rpm, the system knows about all dependencies and warns me if I'm going to hurt them.

Tobias

[Tobias Weiss]

OK, now taking system packages again, the compilation stopped with sagelib again. I'll make a new thread, because its another problem, I think ...

Tobias