build errors on void linux

[Nicolo']

Hi, can anyone help?

I'm trying to build Sage on Void Linux.
My system uses musl and libressl (see also attached list of installed packages).
My goal is to build in a chroot to get a binary package, but for the moment I'm just building locally with

./configure
MAKE='make -j12' make

I get errors building packages, some of which are actually installed on my system.

It seems I cannot attach full install log, which is 27MB, but it ends with

real 7m53.793s

user 31m38.472s

sys 1m31.524s

***************************************************************

Error building Sage.

The following package(s) may have failed to build (not necessarily

during this run of 'make all-start'):

* package:         giac-1.5.0.63-p0

  last build time: Jun 30 20:57

  log file:        /home/nicolo/build/sage-9.1/logs/pkgs/giac-1.5.0.63-p0.log

  build directory: /home/nicolo/build/sage-9.1/local/var/tmp/sage/build/giac-1.5.0.63-p0

* package:         suitesparse-5.6.0

  last build time: Jun 30 20:57

  log file:        /home/nicolo/build/sage-9.1/logs/pkgs/suitesparse-5.6.0.log

  build directory: /home/nicolo/build/sage-9.1/local/var/tmp/sage/build/suitesparse-5.6.0

* package:         ecl-16.1.2.p5

  last build time: Jun 30 20:58

  log file:        /home/nicolo/build/sage-9.1/logs/pkgs/ecl-16.1.2.p5.log

  build directory: /home/nicolo/build/sage-9.1/local/var/tmp/sage/build/ecl-16.1.2.p5

* package:         r-3.6.2.p0

  last build time: Jun 30 20:59

  log file:        /home/nicolo/build/sage-9.1/logs/pkgs/r-3.6.2.p0.log

  build directory: /home/nicolo/build/sage-9.1/local/var/tmp/sage/build/r-3.6.2.p0

[Dima Pasechnik]

Something is very unusual in your setup; e,g. in building R you get

Error in iconv(x, "UTF-8", outputEncoding, sub = "byte", mark = FALSE) :
unsupported conversion from 'UTF-8' to 'ASCII//TRANSLIT'

In building giac, you get even more strange error, a missing gammaf function.
(something to do with musl?)

suitesparse errors out with
gcc: fatal error: cannot read spec file 'libgomp.spec': No such file
or directory
(try to install OpenMP, I guess)

no idea what happens in the ecl build.

> --
> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/a82769f3-5685-4276-8ef6-2782202071f3o%40googlegroups.com.

[Michael Orlitzky]

On 2020-07-01 08:36, Dima Pasechnik wrote:
>
> In building giac, you get even more strange error, a missing gammaf function.
> (something to do with musl?)

Giac upstream probably needs to

#define _GNU_SOURCE

to get the gammaf() function with musl, since it's not standardized.

[Nicolo' Piazzalunga]

Thanks.
I'm puzzled by 'does not support check for system package'/'no suitable
system package' in config.log
E.g. why is it not finding boost, isl, and why is it building ecl (just
to list a few), all of which are installed as system packages?

[Dima Pasechnik]

On Wed, Jul 1, 2020 at 2:23 PM Nicolo' Piazzalunga
<nicolopi...@gmail.com> wrote:
>
> Thanks.
> I'm puzzled by 'does not support check for system package'/'no suitable
> system package' in config.log
> E.g. why is it not finding boost, isl, and why is it building ecl (just
> to list a few), all of which are installed as system packages?

boost and isl should be possible to use from the system (make sure the
headers are installed - on some Linux
distros they are separate packages)

What happened is that apparently you didn't have boost installed when
you ran ./configure 1st time,
which prompted Sage to build its own copy, and now we see

## -------------------------------------------------------------- ##
## Checking whether SageMath should install SPKG boost_cropped... ##
## -------------------------------------------------------------- ##
configure:12971: boost_cropped has already been installed by SageMath
## ------------------------------------------------------ ##
## Checking whether SageMath should install SPKG boost... ##
## ------------------------------------------------------ ##
configure:13439: checking whether any of boost_cropped is installed as
or will be installed as SPKG
configure:13443: result: yes; install boost as well
configure:13500: no suitable system package found for SPKG boost

(boost the Sage package depends on boost_cropped)

>
> On 7/1/20 3:15 PM, Michael Orlitzky wrote:
> > On 2020-07-01 08:36, Dima Pasechnik wrote:
> >>
> >> In building giac, you get even more strange error, a missing gammaf function.
> >> (something to do with musl?)
> >
> > Giac upstream probably needs to
> >
> > #define _GNU_SOURCE
> >
> > to get the gammaf() function with musl, since it's not standardized.
> >
>

> --
> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/6eb2b70d-6946-8d8c-6c66-219fd75634db%40gmail.com.

[Michael Orlitzky]

On 2020-07-01 09:23, Nicolo' Piazzalunga wrote:
> Thanks.
> I'm puzzled by 'does not support check for system package'/'no suitable
> system package' in config.log

We haven't written the script to detect a system copy of ECL yet. We're
in the middle of upgrading to a new major version of it, and only after
that's done will it make sense to detect a system copy (which will have
to be brand new). See,

* https://trac.sagemath.org/ticket/22191
* https://trac.sagemath.org/ticket/29617

> E.g. why is it not finding boost, isl, and why is it building ecl (just
> to list a few), all of which are installed as system packages?

I think Dima answered your question for these two, but in general, you
can check config.log to see what decision the ./configure script made
and why. If you need to dig further, the detection scripts are in

build/pkgs/<name-of-package>/spkg-configure.m4

in the sage source tree.

[Nicolo' Piazzalunga]

I can build R-4.0.2 as standalone, why does it fail in Sage?
As for ecl, perhaps wait until we can use system one?
At a lower priority, config is still not seeing python3, together with

arb cddlib cliquer eclib ecm fflas_ffpack flint flintqs fplll gf2x gfan
givaro glpk gp2c iml isl lcalc libbraiding libgd libsemigroups lrcalc
m4ri m4rie mpc mpfi mpfr mpir nauty ntl pari pari_elldata pari_galdata
pari_galpol pari_nftables pari_seadata pari_seadata_small
perl_cpan_polymake_prereq planarity ppl python3 r rw symmetrica tachyon

[Dima Pasechnik]

On Thu, Jul 2, 2020 at 7:08 PM Nicolo' Piazzalunga
<nicolopi...@gmail.com> wrote:
>
> I can build R-4.0.2 as standalone, why does it fail in Sage?
> As for ecl, perhaps wait until we can use system one?
> At a lower priority, config is still not seeing python3, together with
>
> arb cddlib cliquer eclib ecm fflas_ffpack flint flintqs fplll gf2x gfan
> givaro glpk gp2c iml isl lcalc libbraiding libgd libsemigroups lrcalc
> m4ri m4rie mpc mpfi mpfr mpir nauty ntl pari pari_elldata pari_galdata
> pari_galpol pari_nftables pari_seadata pari_seadata_small
> perl_cpan_polymake_prereq planarity ppl python3 r rw symmetrica tachyon

Could you post what you get in config.log after running ./configure
either on a fresh installation, or after
make distclean
?

>
> On 7/1/20 3:45 PM, Michael Orlitzky wrote:
> > On 2020-07-01 09:23, Nicolo' Piazzalunga wrote:
> >> Thanks.
> >> I'm puzzled by 'does not support check for system package'/'no suitable
> >> system package' in config.log
> >
> > We haven't written the script to detect a system copy of ECL yet. We're
> > in the middle of upgrading to a new major version of it, and only after
> > that's done will it make sense to detect a system copy (which will have
> > to be brand new). See,
> >
> > * https://trac.sagemath.org/ticket/22191
> > * https://trac.sagemath.org/ticket/29617
> >
> >
> >> E.g. why is it not finding boost, isl, and why is it building ecl (just
> >> to list a few), all of which are installed as system packages?
> >
> > I think Dima answered your question for these two, but in general, you
> > can check config.log to see what decision the ./configure script made
> > and why. If you need to dig further, the detection scripts are in
> >
> > build/pkgs/<name-of-package>/spkg-configure.m4
> >
> > in the sage source tree.
> >
>

> --
> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/d26696ae-c476-ba52-6929-2f6f192487c1%40gmail.com.

[Nicolo' Piazzalunga]

Sure,
tar xvf sage-9.1.tar.gz
cd sage-9.1
./configure
gives this.

[Dima Pasechnik]

Your gmp installation is incomplete, missing C++ support,
therefore all the packages that depend on gmp are installed,and not
taken from the system.
And there are dozens such packages, eg mpfr,ntl, arb, flint, pari, eclib...

from your config.log:

configure:10030: checking gmp.h usability
configure:10030: g++ -std=gnu++11 -c -g -O2 conftest.cpp >&5
configure:10030: $? = 0
configure:10030: result: yes
configure:10030: checking gmp.h presence
configure:10030: g++ -E -std=gnu++11 conftest.cpp
configure:10030: $? = 0
configure:10030: result: yes
configure:10030: checking for gmp.h
configure:10030: result: yes
configure:10038: checking gmpxx.h usability
configure:10038: g++ -std=gnu++11 -c -g -O2 conftest.cpp >&5
conftest.cpp:56:10: fatal error: gmpxx.h: No such file or directory

Further, system libgd/gdlib is either not installed, or not well-configured:

configure:22938: $PKG_CONFIG --exists --print-errors "gdlib >= 2.1"
Package gdlib was not found in the pkg-config search path.

similarly, R - it needs libR, maybe a separate component on your system

configure:28542: checking for R
configure:28549: $PKG_CONFIG --exists --print-errors "libR >= $3.4.4"
Package libR was not found in the pkg-config search path.
Perhaps you should add the directory containing `libR.pc'
to the PKG_CONFIG_PATH environment variable

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/9c48c08a-98c7-0ae9-75a1-a373e3f9c3e8%40gmail.com.

[Nicolo' Piazzalunga]

Let me clarify: my problem is that Sage still fails to build (r and ecl
give errors).
As for ecl, I think I understand the problem (a patch is needed for
musl), but could ecl trigger failure of r, which I can however build as
a standalone?

At this stage, I have no issues building stuff that could have been
taken from system.
There are no libgd nor libR available, so it builds them.
Ecl cannot be taken from system at present.
Isl is too old, so it builds it.
Python 3.8.3 is too recent, so it builds it.
I installed gmpxx-devel 6.2.0.

[Dima Pasechnik]

On Fri, 3 Jul 2020, 13:33 Nicolo' Piazzalunga, <nicolopi...@gmail.com> wrote:

Let me clarify: my problem is that Sage still fails to build (r and ecl
give errors).
As for ecl, I think I understand the problem (a patch is needed for
musl), but could ecl trigger failure of r, which I can however build as
a standalone?

you might want to try ecl from

https://trac.sagemath.org/ticket/22191

(by using the branch from that ticket you'd use Sage 9.2.beta2 + that change)

As to R (which is not related to ecl in any way), I really don't know. Something with iconv library.

To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/4efcc301-8778-f16a-43ee-d2706a1b2f56%40gmail.com.

[Matthias Koeppe]

We will need your help to replicate a voidlinux system for testing – see https://trac.sagemath.org/ticket/30044

[Nicolo' Piazzalunga]

Sure, just let me know what to do.
In the meantime, let me note that I can build both R-3.6.2 and R-4.0.2
standalone with ./configure --with-x=no

> --
> You received this message because you are subscribed to a topic in the
> Google Groups "sage-devel" group.
> To unsubscribe from this topic, visit
> https://groups.google.com/d/topic/sage-devel/kiEh8JgbWeQ/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to
> sage-devel+...@googlegroups.com
> <mailto:sage-devel+...@googlegroups.com>.

> To view this discussion on the web visit

> https://groups.google.com/d/msgid/sage-devel/a8663175-7561-4b86-b905-3fd327a153e9o%40googlegroups.com
> <https://groups.google.com/d/msgid/sage-devel/a8663175-7561-4b86-b905-3fd327a153e9o%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Matthias Koeppe]

Edit build/bin/sage-guess-package-system, build/bin/sage-print-system-package-command, build/bin/write-dockerfile.sh, tox.ini, and add voidlinux.txt files in build/pkgs/

similar to https://git.sagemath.org/sage.git/commit?id=eecd68a1a3cd6046a6ae0d4b5eb57895ee34a1ba

More info: https://doc.sagemath.org/html/en/developer/portability_testing.html#testing-sage-on-a-different-platform-using-docker

> sage-...@googlegroups.com
> <mailto:sage-devel+unsub...@googlegroups.com>.

[Dima Pasechnik]

On Fri, Jul 3, 2020 at 2:34 PM Nicolo' Piazzalunga
<nicolopi...@gmail.com> wrote:
>
> Sure, just let me know what to do.
> In the meantime, let me note that I can build both R-3.6.2 and R-4.0.2
> standalone with ./configure --with-x=no

you can go to the directory where R was being built by Sage
(where exactly you can see in logs/pkgs/r-3*.log file, at the end)
and inspect the parameters to configure used. Then you can try to
build R with these parameters.
Specifically, look into config.log, or run

./config.status --config

> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/26f36654-2362-59be-346c-e80fabb7bca1%40gmail.com.

[Nicolo' Piazzalunga]

I tried to edit relevant files, see diff.log attached.
Please check docker-specific files, as I'm not familiar with it.
(The relevant image in docker-hub is /voidlinux/masterdir-x86_64-musl)
I have not yet dealt with equivalent distribution packages, if it's ok I
will do it once we have no build errors.)
Notice that it uses musl libc and libressl (instead of openssl).
I attach the list of packages installed on the machine I'm building on.

[Matthias Koeppe]

Thanks. I have put your changes on the branch of https://trac.sagemath.org/ticket/30044 and did some additional changes.

You can now test using:

tox -e docker-voidlinux-standard

Currently, it fails as follows:

src/doc/bootstrap:85: installing src/doc/en/reference/repl/options.txt

gettext and the gettextize program must be installed and be in

your PATH. E.g. Homebrew installs them in /usr/local/opt/gettext/bin.

Bootstrap failed. Either upgrade autotools or run bootstrap with

the -d option to download the auto-generated files instead.

This means that some package are missing.

[Nicolo' Piazzalunga]

Does adding
libltdl-devel python3-pkgconfig gettext-devel
help?

[Matthias Koeppe]

Thanks. I have added gettext-devel to build/pkgs/void-bootstrap.txt on the ticket.

"tox -e docker-voidlinux-standard" is now beginning to build Sage. I'll let you know if it reproduces the errors that you reported.

[Matthias Koeppe]

I can already confirm a build error in the ecl package.

You may want to try with ticket https://trac.sagemath.org/ticket/22191 , which makes a major update to ECL.

[Matthias Koeppe]

Also the "giac" build error is confirmed.

On the other hand "r" and "suitesparse" build successfully.

To reproduce the build errors that you reported, information about equivalent distribution packages needs to be added to build/pkgs/*/distros/void.txt

[Nicolo' Piazzalunga]

Let us include giac-devel (1.5.0.87):

diff --git a/build/pkgs/giac/distros/void.txt
b/build/pkgs/giac/distros/void.txt
new file mode 100644
index 0000000000..daf8fd8679
--- /dev/null
+++ b/build/pkgs/giac/distros/void.txt
@@ -0,0 +1,2 @@
+giac-devel
+libgiac

It seems ecl (16.1.3) needs to be patched, as in
https://github.com/void-linux/void-packages/tree/master/srcpkgs/ecl
Shall we test with version 20 or try to work with 16?

> --
> You received this message because you are subscribed to a topic in the
> Google Groups "sage-devel" group.
> To unsubscribe from this topic, visit
> https://groups.google.com/d/topic/sage-devel/kiEh8JgbWeQ/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to

> sage-devel+...@googlegroups.com
> <mailto:sage-devel+...@googlegroups.com>.

> To view this discussion on the web visit

> https://groups.google.com/d/msgid/sage-devel/43be6f39-f9cc-4738-887a-00c2bea3a988o%40googlegroups.com
> <https://groups.google.com/d/msgid/sage-devel/43be6f39-f9cc-4738-887a-00c2bea3a988o%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Dima Pasechnik]

the ecl 20 ticket has been positively reviewed. there is no need to spend time fighting problems caused by the outdated ecl 16.

You received this message because you are subscribed to the Google Groups "sage-devel" group.

To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/ab215f6e-7552-0184-4e65-14ec74f11ddf%40gmail.com.

[Matthias Koeppe]

On Friday, July 10, 2020 at 6:55:06 AM UTC-7, Nicolo' wrote:

diff --git a/build/pkgs/giac/distros/void.txt
b/build/pkgs/giac/distros/void.txt
new file mode 100644
index 0000000000..daf8fd8679
--- /dev/null
+++ b/build/pkgs/giac/distros/void.txt
@@ -0,0 +1,2 @@
+giac-devel
+libgiac

Thanks. More packages, please 

[Nicolo' Piazzalunga]

With this, we should get close to the list of explicitly installed
packages on my machine (attached).

On 7/10/20 6:18 PM, Matthias Koeppe wrote:
>
> Thanks. More packages, please
>

[Matthias Koeppe]

Thank you. I have added it to the branch. 

By the way, I noticed that there is no "latest" tag on the Docker image. Perhaps the voidlinux people could add this.

Also, for sending patches here in the group, it's best to use "git commit" followed by "git format-patch".

[Matthias Koeppe]

I am getting the following error:

Aborting!

Package `which' already installed.

Unable to locate 'glib2' in repository pool.

The command '/bin/sh -c xbps-install -Su && xbps-install --yes binutils make m4 perl python3 tar bc gcc which gettext autoconf automake libtool pkg-config gettext-devel bzip2-devel curl libcurl-devel ecm-devel flint freetype harfbuzz glib glib2 glpk-devel gmpxx-devel gsl-devel imlib2-devel mpc libmpc-devel mpfr ncurses-devel openblas-devel pcre-devel pcre2-devel ppl-devel python3 python3-devel readline-devel rw sqlite-devel yasm zeromq-devel' returned a non-zero code: 2

[Nicolo' Piazzalunga]

Sorry, that was likely a typo from me.
Does it work now? is the format ok?

From bfd28b43c42cf2f24427f9ac0fa0c4bcd8dc475e Mon Sep 17 00:00:00 2001
From: nicolo <nicolo@...>
Date: Wed, 22 Jul 2020 15:46:41 +0200
Subject: [PATCH] remove glib2 (typo?)

---
build/pkgs/freetype/distros/void.txt | 1 -
1 file changed, 1 deletion(-)

diff --git a/build/pkgs/freetype/distros/void.txt
b/build/pkgs/freetype/distros/void.txt
index d9f02e6c56..8ca004ddd3 100644
--- a/build/pkgs/freetype/distros/void.txt
+++ b/build/pkgs/freetype/distros/void.txt
@@ -1,4 +1,3 @@
freetype
harfbuzz
glib
-glib2
--
2.27.0

[Matthias Koeppe]

Thank you. I have added it to the branch.

It's better to post these files as an attachment - as inline text it gets mangled by Google Groups.

Also, see https://doc.sagemath.org/html/en/developer/git_setup.html#your-name-and-email 

I'll run another round of tests with "tox -e docker-voidlinux-standard" now.

[Nicolo' Piazzalunga]

> By the way, I noticed that there is no "latest" tag on the Docker image.
> Perhaps the voidlinux people could add this.

Note that Void Linux is a rolling release.

[Matthias Koeppe]

Exactly -- right now I am forced to hardcode voidlinux/masterdir-x86_64-musl:20200104 in tox.ini because there is no "latest" tag...

[Matthias Koeppe]

On Wednesday, July 22, 2020 at 10:21:31 AM UTC-7, Matthias Koeppe wrote:

I'll run another round of tests with "tox -e docker-voidlinux-standard" now.

It reproduces the build failure of giac-1.5.0.63-p0 that you reported. 

Note that although you added the giac system package information, the system package is not used yet because there is no spkg-configure.m4 script for giac.

Several ways forward:

- https://trac.sagemath.org/ticket/29541 spkg-configure.m4 for giac

- https://trac.sagemath.org/ticket/29552 Upgrade giac to 1.5.0-87 (or 1.6.x)

[Nicolo' Piazzalunga]

Thanks.
Let me look into giac.
Still no errors with R? Does it mean it is built, or it fails before it
tries to build it?

[Matthias Koeppe]

R builds successfully.

[Nicolo' Piazzalunga]

I have added pari, which I either previously overlooked or has been
recently added.

Void people should be taking care of renaming docker images.

Giac is already at 1.5.0.87, so perhaps try the new spkg-configure.m4?

[Nicolo' Piazzalunga]

On 8/17/20 1:34 PM, Nicolo' Piazzalunga wrote:
>
> Giac is already at 1.5.0.87, so perhaps try the new spkg-configure.m4?
>

Indeed if I try to build /sage-9.2.beta8, it only fails with
giac-1.5.0.63-p0

[Matthias Koeppe]

On Monday, August 17, 2020 at 4:34:51 AM UTC-7, Nicolo' wrote:

I have added pari, which I either previously overlooked or has been
recently added.

Thanks. I have put it on https://trac.sagemath.org/ticket/30382

 

Void people should be taking care of renaming docker images.

Great, thanks for contacting them.

 

Giac is already at 1.5.0.87, so perhaps try the new spkg-configure.m4?

Yes, https://trac.sagemath.org/ticket/29541 is worth trying and needs review

 

[Nicolo' Piazzalunga]

After importing
u/mkoeppe/spkg_configure_m4_for_giac
u/mkoeppe/more_voidlinux_txt
and d43730c (should this appear in beta9?),
I still get

the following SPKGs did not find equivalent system packages: arb brial
cddlib cliquer coxeter3 eclib fflas_ffpack flint flintqs fplll gf2x gfan
givaro gp2c iml isl lcalc libbraiding libgd libsemigroups lrcalc m4ri
m4rie mpfi nauty ntl palp pari pari_elldata pari_galdata pari_galpol
pari_nftables pari_seadata pari_seadata_small perl_cpan_polymake_prereq
planarity python3 r rw suitesparse symmetrica sympow tachyon zn_poly

checking for the package system in use... void
configure: hint: installing the following system packages is recommended
and may avoid building some of the above SPKGs from source:
configure: $ sudo xbps-install flint imlib2-devel isl16-devel pari
pari-devel python3 python3-devel rw

But these are already installed (notice there's no isl16-devel, maybe
isl15-devel?). In particular,

pari-2.11.4.p0: no suitable system package;
will be installed as an SPKG

giac-1.5.0.87: does not support check for
system package; will be installed as an SPKG

python3-3.7.3.p1: no suitable system package;
will be installed as an SPKG

Shouldn't all these be detected by now?

[Matthias Koeppe]

config.log has more information on why packages are rejected

[Nicolo' Piazzalunga]

Thanks, I figured out almost all my issues.
Except, only mention of giac in config.log is

giac-1.5.0.87: does not support check for system package

Shouldn't it be able to check it by now?
I attach a patch for R and isl15-devel.

> --
> You received this message because you are subscribed to a topic in the
> Google Groups "sage-devel" group.
> To unsubscribe from this topic, visit
> https://groups.google.com/d/topic/sage-devel/kiEh8JgbWeQ/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to
> sage-devel+...@googlegroups.com
> <mailto:sage-devel+...@googlegroups.com>.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/sage-devel/3ee29877-7a7f-46a3-8536-a73aae44773ao%40googlegroups.com
> <https://groups.google.com/d/msgid/sage-devel/3ee29877-7a7f-46a3-8536-a73aae44773ao%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Matthias Koeppe]

Try again after "./bootstrap" (this needs autotools installed)

> sage-...@googlegroups.com
> <mailto:sage-devel+unsub...@googlegroups.com>.

[Nicolo' Piazzalunga]

On 8/18/20 5:03 PM, Matthias Koeppe wrote:
> Try again after "./bootstrap" (this needs autotools installed)

Thanks, now it checks, but since pari misses some parts, it doesn't
detect giac either.

## ----------------------------------------------------- ##
## Checking whether SageMath should install SPKG giac... ##
## ----------------------------------------------------- ##
configure:20077: checking whether any of pari is installed as or will be
installed as SPKG
configure:20081: result: yes; install giac as well
configure:20296: no suitable system package found for SPKG giac

## ----------------------------------------------------- ##
## Checking whether SageMath should install SPKG pari... ##
## ----------------------------------------------------- ##
configure:17312: checking whether any of gmp mpir readline is installed
as or will be installed as SPKG
configure:17321: result: no
configure:17326: checking for gp
configure:17344: found /bin/gp
configure:17356: result: /bin/gp
configure:17370: checking for gphelp
configure:17388: found /bin/gphelp
configure:17400: result: /bin/gphelp
configure:17417: checking whether gphelp has access to the documentation
configure:17420: result: yes
configure:17432: checking is pari_elldata installed?
*** at top-level: r=ellinit("11a1");r[11]
*** ^---------------------
*** ellinit: error opening elldata file: `/usr/share/pari/elldata/ell0'.
configure:17439: result: no; cannot use system pari/GP without elldata
package
configure:17441: Install elldata package and reconfigure.
configure:17443: Otherwise Sage will build its own pari/GP.
configure:17447: checking is pari_galdata installed?
*** at top-level: polgalois(x^8-2)[1]
*** ^-------------------
*** polgalois: error opening galois file:
`/usr/share/pari/galdata/COS8_50_47'.
configure:17454: result: no; cannot use system pari/GP without galdata
package
configure:17456: Install galdata package and reconfigure.
configure:17458: Otherwise Sage will build its own pari/GP.
configure:17462: checking is pari_galpol installed?
*** at top-level: galoisgetname(12,1)=="C3 : C4"
*** ^------------------------------
*** galoisgetname: error opening galpol file:
`/usr/share/pari/galpol/12/nb'.
configure:17469: result: no; cannot use system pari/GP without galpol
package
configure:17471: Install galpol package and reconfigure.
configure:17473: Otherwise Sage will build its own pari/GP.
configure:17477: checking is pari_seadata installed?
*** at top-level: poldegree(ellmodulareqn(211)[1])
*** ^----------------------
*** ellmodulareqn: error opening seadata file:
`/usr/share/pari/seadata/sea211'.
configure:17484: result: no; cannot use system pari/GP without seadata
package
configure:17486: Install seadata package and reconfigure.
configure:17488: Otherwise Sage will build its own pari/GP.
configure:17492: checking whether hyperellcharpoly bug is fixed
configure:17497: result: yes
configure:17508: checking whether bnfisunit bug of pari 2.11.3 is fixed
configure:17513: result: yes
configure:17750: no suitable system package found for SPKG pari

[Dima Pasechnik]

On Wed, Aug 19, 2020 at 9:47 AM Nicolo' Piazzalunga
<nicolopi...@gmail.com> wrote:
>
> On 8/18/20 5:03 PM, Matthias Koeppe wrote:
> > Try again after "./bootstrap" (this needs autotools installed)
>
> Thanks, now it checks, but since pari misses some parts, it doesn't
> detect giac either.

these missing pari/gp parts are just data, easy to package.

> --
> You received this message because you are subscribed to the Google Groups "sage-devel" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/72f51a7a-f860-d404-7272-872352677738%40gmail.com.

[Nicolo' Piazzalunga]

> these missing pari/gp parts are just data, easy to packageGood, working on that.
In the meantime, it is useful to add a few pkgs to bootstrap (attached).

[Dima Pasechnik]

Please open a trac ticket on this, and, ideally, create a branch with
this change.

>
> --
> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/cb342d98-a27e-4b76-c105-ef98ed5bd8c6%40gmail.com.

[Matthias Koeppe]

Thanks. I've pushed this to https://trac.sagemath.org/ticket/30382

[Nicolo' Piazzalunga]

Finally sage builds successfully on void linux :)
I will add spkg files for pari-seadata etc as soon as the packages are
approved, this way it does not have to build pari and giac.

Is there a ticket already to use python3.8 from system?

Should I be worried that it fails some doctest?

sage -t --random-seed=0 src/sage/interfaces/interface.py # 10 doctests
failed
sage -t --random-seed=0 src/sage/interfaces/r.py # 191 doctests failed
sage -t --random-seed=0 src/sage/interfaces/giac.py # 2 doctests failed
sage -t --random-seed=0 src/sage/symbolic/expression.pyx # 1 doctest failed
sage -t --random-seed=0 src/sage/lfunctions/sympow.py # 3 doctests failed
sage -t --random-seed=0 src/sage/stats/r.py # 1 doctest failed
sage -t --random-seed=0 src/sage/misc/sageinspect.py # 1 doctest failed
sage -t --random-seed=0 src/sage/misc/gperftools.py # 1 doctest failed
sage -t --random-seed=0 src/sage/repl/ipython_tests.py # 2 doctests failed
sage -t --random-seed=0
src/doc/en/prep/Quickstarts/Statistics-and-Distributions.rst # 6
doctests failed

Thanks.

[Samuel Lelievre]

2020-08-21 12:40:44 UTC, Nicolo':

>

> Is there a ticket already to use python3.8 from system? 

The ticket to upgrade Sage-the-distribution to Python 3.8

just got positive review:

  https://trac.sagemath.org/ticket/27754

By clicking on the link in the "Branch" field one can see

that it changes `build/pkgs/python3/spkg-configure.m4`

to accept Python >= 3.6 and < 3.9.

Using Trac's (case-insensitive) query engine, search for tickets

with "3.8" in the summary sorted by decreasing ticket number:

  https://trac.sagemath.org/query?order=id&desc=1&summary=~3.8

Tip: most browsers let you define "search engine shortcuts".

In my browser I defined "tsum" (for "ticket summary") as

  https://trac.sagemath.org/query?order=id&desc=1&summary=~%s

so that I can type this in the url bar

  tsum 3.8

and it expands to the above. Similarly "tdesc", "tkey, "tcc", etc.

let me search in the "description", "keywords", "cc", etc. fields,

and "nc" variants search for non-closed tickets by adding

    &status=!closed

into the search query.

[Nicolo' Piazzalunga]

On 8/21/20 7:19 PM, Samuel Lelievre wrote:
>
> The ticket to upgrade Sage-the-distribution to Python 3.8
> just got positive review:
>
> https://trac.sagemath.org/ticket/27754
>

Thanks. I also give it a positive review, hopefully it'll be merged
soon. Now on my Void machine sage builds in 30m (previously 1h 30m).


As for doctest errors, it's as before, except misc/gperftools.py failure
disappeared. Is it to be expected to get these failures when building
beta8 version?

[Nicolo' Piazzalunga]

Would it be possible to create a new trac ticket with more void spkg's?
I attach a patch for pari subpackages (waiting to be merged in void).
Leaving it open for a while will simplify adding spkg's and building.

Thanks.

[Dima Pasechnik]

On Tue, Sep 8, 2020 at 4:32 PM Nicolo' Piazzalunga
<nicolopi...@gmail.com> wrote:
>
> Would it be possible to create a new trac ticket with more void spkg's?

sure - can you do this yourself?

> I attach a patch for pari subpackages (waiting to be merged in void).
> Leaving it open for a while will simplify adding spkg's and building.
>
> Thanks.
>

> --
> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/f750e2fa-c832-62eb-c9aa-4ae551704887%40gmail.com.

[Nicolo' Piazzalunga]

I did it.

On 9/8/20 5:39 PM, Dima Pasechnik wrote:
> sure - can you do this yourself?

It seems sage is still unable to detect my pari-seadata package, even
though the file /usr/share/pari/seadata/sea211 is present on my system.

Do you know what the problem could be?


configure:17484: checking is pari_seadata installed?

*** at top-level: poldegree(ellmodulareqn(211)[1])
*** ^----------------------
*** ellmodulareqn: error opening seadata file:
`/usr/share/pari/seadata/sea211'.

gp_seadat_check=`echo "poldegree(ellmodulareqn(211)[1])" | $GP -qf 2>>
config.log`
if test x$gp_seadat_check = x212; then
{ $as_echo "$as_me:${as_lineno-$LINENO}: result: yes" >&5
$as_echo "yes" >&6; }


Thanks.

[Dima Pasechnik]

On Tue, Sep 15, 2020 at 2:11 PM Nicolo' Piazzalunga <nicolopi...@gmail.com> wrote:

I did it.

On 9/8/20 5:39 PM, Dima Pasechnik wrote:
> sure - can you do this yourself?

It seems sage is still unable to detect my pari-seadata package, even
though the file /usr/share/pari/seadata/sea211 is present on my system.

Do you know what the problem could be?

I think it looks at Pari/GP datadir to locate these files. What's its value?

At GP prompt, type

? default(datadir)

and see (e.g.)

 

%1 = "/usr/local/share/pari" 

in this case GP expects seadata to be in

/usr/local/share/pari/seadata/

HTH

Dima

 configure:17484: checking is pari_seadata installed?
   ***   at top-level: poldegree(ellmodulareqn(211)[1])
   ***                           ^----------------------
   *** ellmodulareqn: error opening seadata file:
`/usr/share/pari/seadata/sea211'.


gp_seadat_check=`echo "poldegree(ellmodulareqn(211)[1])" | $GP -qf 2>>
config.log`
if test x$gp_seadat_check = x212; then
         { $as_echo "$as_me:${as_lineno-$LINENO}: result: yes" >&5
$as_echo "yes" >&6; }


Thanks.

--
You received this message because you are subscribed to the Google Groups "sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/d2a4b9ff-9765-e68b-2f83-f63216dd752c%40gmail.com.

[Nicolo' Piazzalunga]

Thanks. I get:

? default(datadir)
%1 = "/usr/share/pari"

so it seems it is correctly configured, but still not detected..

[Dima Pasechnik]

If you are trying this on an existing install, did you run

make pari-clean

before trying ./configure ?



Please send the top-level config.log

> --
> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/56132297-ac87-c675-72f1-92c33a74c5f3%40gmail.com.

[Nicolo' Piazzalunga]

It was on a fresh install, see the attached config.log

[Dima Pasechnik]

On Wed, Sep 16, 2020 at 1:33 PM Nicolo' Piazzalunga
<nicolopi...@gmail.com> wrote:
>
> It was on a fresh install, see the attached config.log
>

does

poldegree(ellmodulareqn(211)[1])

work at the gp prompt, interactively (with the same user id as used
for running ./configure)
(more precisely, your gp is found in /bin/, so you might want to check
that /bin/gp and gp are the same
thing if run from the shell)

I guess it might be wrong file permissions for a part of
/usr/share/pari/ - as tests for other data packages there succeed

configure:17333: checking for gp
configure:17351: found /bin/gp
configure:17363: result: /bin/gp
configure:17377: checking for gphelp
configure:17395: found /bin/gphelp
configure:17407: result: /bin/gphelp
configure:17424: checking whether gphelp has access to the documentation
configure:17427: result: yes
configure:17439: checking is pari_elldata installed?
configure:17443: result: yes
configure:17454: checking is pari_galdata installed?
configure:17458: result: yes
configure:17469: checking is pari_galpol installed?
configure:17473: result: yes

configure:17484: checking is pari_seadata installed?
*** at top-level: poldegree(ellmodulareqn(211)[1])
*** ^----------------------
*** ellmodulareqn: error opening seadata file:
`/usr/share/pari/seadata/sea211'.

What are the permissions of /usr/share/pari/seadata/sea211 ?
(i.e. the output of
ls -l /usr/share/pari/seadata/sea211
please)

> On 9/16/20 1:03 PM, Dima Pasechnik wrote:
> > If you are trying this on an existing install, did you run
> >
> > make pari-clean
> >
> > before trying ./configure ?
> >
> >
> >
> > Please send the top-level config.log
>

> --
> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/5289ca29-b71b-883c-c437-aac312a98127%40gmail.com.

[Nicolo' Piazzalunga]

Same from gp prompt:

? poldegree(ellmodulareqn(211)[1])

*** at top-level: poldegree(ellmodulareqn(211)[1])
*** ^----------------------
*** ellmodulareqn: error opening seadata file:
`/usr/share/pari/seadata/sea211'.

*** Break loop: type 'break' to go back to GP prompt

You're right, there's a permission issue:

ls -l /usr/share/pari/seadata/sea211
-rw-r--r-- 1 root root 194468 Sep 15 12:01 /usr/share/pari/seadata/sea211

Thanks.

[Nicolo' Piazzalunga]

Problem seems now solved.
I updated ticket #30539, where I included extra pari spkgs.

Thanks.