Re: [sage-release] Sage 9.2.beta12 released

[Samuel Lelievre]

2020-09-11 22:31:21 UTC, Dima Pasechnik:
>
> On Fri, Sep 11, 2020 at 11:16 PM Samuel Lelievre:
>>
>> I'll reinstall gcc@9 and python@3.7 and  try again.
>
> but why, install gcc (i.g. gcc@10), Sage now accepts it.
> Probably you want to configure with CC=clang CXX=clang++
>
> You may also try installing python3.8 and use it
> with the experimental branch from
> https://trac.sagemath.org/ticket/30546

Can "git trac try" take several ticket numbers so as
to apply them both on top of the latest develop?
For instance will this work?

$ git trac try 30536 30546

Otherwise hopefully the next beta includes
at least 30536 which has positive review.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/4efba0c6-1a2d-4db0-83d4-a216b322ff8co%40googlegroups.com.

Re: [sage-release] Sage 9.2.beta12 released

[Matthias Köppe]

On Friday, September 11, 2020 at 3:31:21 PM UTC-7, Dima Pasechnik wrote:
>
> Probably you want to configure with CC=clang CXX=clang++
>

That's actually not necessary. homebrew does not provide unversioned gcc 
and g++ executables, so gcc ->  /usr/bin/gcc and g++ -> /usr/bin/g++

 

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/45750fc3-9ed1-4d0a-b56e-c6649c1aab7do%40googlegroups.com.

Re: [sage-release] Sage 9.2.beta12 released

[Dima Pasechnik]

On Fri, Sep 11, 2020 at 11:16 PM Samuel Lelievre 
wrote:

>
>
> Le vendredi 11 septembre 2020 05:55:47 UTC+2, Samuel Lelievre a écrit :
>>
>>
>>
>> Le jeudi 10 septembre 2020 19:15:43 UTC+2, Matthias Köppe a écrit :
>>>
>>> On Thursday, September 10, 2020 at 10:08:32 AM UTC-7, Samuel Lelievre
>>> wrote:

 On a MacBook Air (early 2014) with
 macOS 10.14.6 Mojave, with Homebrew
 (but having removed Homebrew packages
 gcc@9, python@3.7, arb, flint, gap, singular),
 `make` succeeds but `make testlong` fails
 unexpectedly still referring to python@3.7.

>>>
>>> "make distclean" and rebuilding should fix this
>>>
>>
>> I forgot to mention this was a fresh git clone.
>> I still tried your suggestion but testing failures
>> persist after "make distclean" and rebuilding.
>>
>
> I'll reinstall gcc@9 and python@3.7 and  try again.
>

but why, install gcc (i.g. gcc@10), Sage now accepts it.
Probably you want to configure with CC=clang CXX=clang++

 You may also try installing python3.8 and use it with the experimental
branch from https://trac.sagemath.org/ticket/30546

> --
> You received this message because you are subscribed to the Google Groups
> "sage-release" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sage-release+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sage-release/43d187fc-34aa-409d-bde2-caef240e5860o%40googlegroups.com
> 
> .
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/CAAWYfq0QxQGK9tEiREbvpJBDHx1fPhNTLCJoDzDu2x%3D23FQmHA%40mail.gmail.com.

Re: [sage-release] Sage 9.2.beta12 released

[Samuel Lelievre]

Le vendredi 11 septembre 2020 05:55:47 UTC+2, Samuel Lelievre a écrit :
>
>
>
> Le jeudi 10 septembre 2020 19:15:43 UTC+2, Matthias Köppe a écrit :
>>
>> On Thursday, September 10, 2020 at 10:08:32 AM UTC-7, Samuel Lelievre 
>> wrote:
>>>
>>> On a MacBook Air (early 2014) with 
>>> macOS 10.14.6 Mojave, with Homebrew 
>>> (but having removed Homebrew packages 
>>> gcc@9, python@3.7, arb, flint, gap, singular), 
>>> `make` succeeds but `make testlong` fails 
>>> unexpectedly still referring to python@3.7. 
>>>
>>
>> "make distclean" and rebuilding should fix this
>>
>
> I forgot to mention this was a fresh git clone.
> I still tried your suggestion but testing failures
> persist after "make distclean" and rebuilding.
>

I'll reinstall gcc@9 and python@3.7 and  try again.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/43d187fc-34aa-409d-bde2-caef240e5860o%40googlegroups.com.

Re: [sage-release] Re: Sage 9.2.beta12 released

[Matthias Köppe]

On Friday, September 11, 2020 at 12:06:51 PM UTC-7, Markus Wageringel wrote:
>
> I think it might be better to stop the build earlier (and maybe ask for 
> confirmation) before proceeding to build Python, to avoid building Python 
> and all its dependencies a second time after installing OpenSSL. That is, 
> unless there is a reason to build Sage without OpenSSL. 
>

After the license change that will come with OpenSSL 3 
(https://trac.sagemath.org/ticket/29555), we can make openssl a standard 
package. 

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/a3300178-664e-4cd8-807a-f90b29ce4aefo%40googlegroups.com.

Re: [sage-release] Re: Sage 9.2.beta12 released

[Markus Wageringel]

> Am 11.09.2020 um 20:46 schrieb Samuel Lelièvre :
> 
> On Debian or Debian-based system like Ubuntu, check whether
> libssl-dev is installed with this command (sudo not needed):
> ```
> $ dpkg-query --list libssl-dev
> ```

Thank you for the explanations. The command returns:

$ dpkg-query --list libssl-dev
dpkg-query: no packages found matching libssl-dev

Also, the directory Matthias mentioned `/usr/include/openssl` does not exist.


> It would be very useful when building on a system
> without the appropriate ssl dependencies like yours
> or most macOS systems to end the build with a word
> such as
> 
>  Sage was successfully built, and starts,
>  but be warned, it build its own Python
>  without the SSL module. To know more, visit
>

I think it might be better to stop the build earlier (and maybe ask for 
confirmation) before proceeding to build Python, to avoid building Python and 
all its dependencies a second time after installing OpenSSL. That is, unless 
there is a reason to build Sage without OpenSSL.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/1D005FF4-9D1F-4735-802B-26AAAB26D507%40gmail.com.

[sage-release] Re: Sage 9.2.beta12 released

[Antonio Rojas]

This is the issue I mentioned 
in https://trac.sagemath.org/ticket/29240#comment:3
For some time I could reproduce it on sage-the-distro but not on my Arch 
package, then I could no longer reproduce it on sage-the-distro and 
suddenly started happening on the Arch package, triggered by the sympy 
1.6.1 update (which shuffles some imports).
It seems that the singular instance is being unloaded from memory, which 
causes pexpect to crash. While trying to debug this, I found that removing 
almost *any* test block from singular.py will prevent it from crashing, 
which is certainly weird.

El viernes, 11 de septiembre de 2020 a las 20:15:50 UTC+2, 
matthia...@gmail.com escribió:

> On Friday, September 11, 2020 at 10:55:57 AM UTC-7, Markus Wageringel 
> wrote:
>>
>> 9.2.beta11 has introduced crashes of the doctests in 
>> `src/sage/interfaces/singular.py` that were reproducible on about 20% of 
>> the test runs and have been observed by several other people as well. I 
>> have tracked that issue down to the Pexpect upgrade in #29240 via 
>> bisection. With 9.2.beta12 on the other hand, I cannot reproduce the 
>> problem anymore, possibly because of the Python upgrade. There are still 
>> reports about this with 9.2.beta12 though, so maybe this helps analyze the 
>> problem.
>>
>
> Is singular crashing in these tests or Sage? If it's singular, it would be 
> helpful to enable logging in the Pexpect interface and then trying to 
> reproduce the errors without Sage
>
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/b74f094b-f90a-41c4-83ef-91bf6ca3ff5bn%40googlegroups.com.

Re: [sage-release] Re: Sage 9.2.beta12 released

[Samuel Lelièvre]

2020-09-11 17:55 UTC, Markus Wageringel:
>
> With a clean build on Ubuntu 20.04, ptestlong passes, but Jupyter
> does not start anymore due to an SSL problem, using Sage's Python 3.
> (The system Python 3.8 is not picked up because of missing
> dependencies sqlite and xz, but that is ok.)
>
> $ ./sage -n jupyter
> Please wait while the Sage Jupyter Notebook server starts...
> Traceback (most recent call last):
> ...
> import _ssl # if we can't import it, let the error 
> propagate
> ModuleNotFoundError: No module named '_ssl'
>
> The Jupyter notebook requires ssl, even if you do not use
> https. Install the openssl development packages in your system and
> then rebuild Python (sage -f python3).
>
> Which packages need to be installed for this and why was this not
> a problem previously? OpenSSL is installed on the system:
>
> $ openssl version
> OpenSSL 1.1.1f  31 Mar 2020

The build system apparently did not build Python's
`_ssl` module when it built Python. It is supposed to
always build it provided openssl is present, either from
the system or built as an spkg.

Or could it be that your system has OpenSSL but not
libssl-dev, as Matthias suggests?

2020-09-11 18:17 UTC, Matthias Köppe:
>
> Is "libssl-dev" installed?
> (see build/pkgs/openssl/distros/debian.txt)

On Debian or Debian-based system like Ubuntu, check whether
libssl-dev is installed with this command (sudo not needed):
```
$ dpkg-query --list libssl-dev
```

If indeed it was not installed, do this:
```
$ sudo apt install libssl-dev
$ ./sage -f python3
```

Oh, but you replied to Matthias now:

2020-09-11 18:34 UTC, Markus Wageringel
>
> It looks like it is not:

In the meantime I had opened a ticket:

- Build Python's ssl module
  https://trac.sagemath.org/ticket/30556

in which I forgot about the libssl vs libssl-dev subtlety.

I'll change the ticket's goal to

- Warn when building Python without its SSL module

I know when there are errors, they are summarized
at the end of the build. Do we have a mechanism
to collect warnings and summarise them at the end
of the build?

It would be very useful when building on a system
without the appropriate ssl dependencies like yours
or most macOS systems to end the build with a word
such as

  Sage was successfully built, and starts,
  but be warned, it build its own Python
  without the SSL module. To know more, visit


Would there be a way to make `sage -n jupyter` first check
for the ssl module, and if it is not found provide some
helpful error message with hints as to how to fix things?

If you get a chance to ask someone with sudo rights
on the machine you were using, have them install all
the pieces you were missing: xz, sqlite, libssl-dev,
for next time!

See at the end of the output of `./configure` for a list
of recommended packages to install system-wide.

In the meantime a workaround as you guessed is to build
Sage's openssl spkg and rebuild Sage's Python so it takes
advantage of it:
```
$ ./sage -i openssl
$ ./sage -f python3
```

This last command will also rebuild any packages
that depend on the `python3` spkg.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/CAEcArF1hjuRS0u982FznHqjQcN6aopJHmAbmHbs44yHrTbgcpw%40mail.gmail.com.

Re: [sage-release] Re: Sage 9.2.beta12 released

[Matthias Köppe]

On Friday, September 11, 2020 at 11:34:15 AM UTC-7, Markus Wageringel wrote:
>
> > Am 11.09.2020 um 20:17 schrieb Matthias Köppe  >: 
> > 
> > Is "libssl-dev" installed? (see build/pkgs/openssl/distros/debian.txt) 
>
>
(This is the name of the Debian package with the header files.) Check for 
files in /usr/include/openssl

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/8dc733e8-21b2-4641-9509-5568f724a205o%40googlegroups.com.

Re: [sage-release] Re: Sage 9.2.beta12 released

[Markus Wageringel]

> Am 11.09.2020 um 20:17 schrieb Matthias Köppe :
> 
> Is "libssl-dev" installed? (see build/pkgs/openssl/distros/debian.txt)

It looks like it is not:

$ ldconfig -p | grep ssl
libxmlsec1-openssl.so.1 (libc6,x86-64) => 
/lib/x86_64-linux-gnu/libxmlsec1-openssl.so.1
libssl3.so (libc6,x86-64) => /lib/x86_64-linux-gnu/libssl3.so
libssl.so.1.1 (libc6,x86-64) => /lib/x86_64-linux-gnu/libssl.so.1.1
libgnutls-openssl.so.27 (libc6,x86-64) => 
/lib/x86_64-linux-gnu/libgnutls-openssl.so.27

I do not have sudo rights to install it. Enabling the optional OpenSSL SPKG 
would fix this then, I assume.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/A5126C01-F5D6-4DEE-848A-029BA7495201%40gmail.com.

[sage-release] Re: Sage 9.2.beta12 released

[Matthias Köppe]

On Friday, September 11, 2020 at 10:55:57 AM UTC-7, Markus Wageringel wrote:
>
> With a clean build on Ubuntu 20.04, ptestlong passes, but Jupyter does not 
> start anymore due to an SSL problem, Which packages need to be installed 
> for this and why was this not a problem previously? OpenSSL is installed on 
> the system:
>
> $ openssl version
> OpenSSL 1.1.1f  31 Mar 2020
>
>
Is "libssl-dev" installed? (see build/pkgs/openssl/distros/debian.txt)
 

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/cbf8507e-6691-4b32-b26f-c4e02d115801o%40googlegroups.com.

[sage-release] Re: Sage 9.2.beta12 released

[Matthias Köppe]

On Friday, September 11, 2020 at 10:55:57 AM UTC-7, Markus Wageringel wrote:
>
> 9.2.beta11 has introduced crashes of the doctests in 
> `src/sage/interfaces/singular.py` that were reproducible on about 20% of 
> the test runs and have been observed by several other people as well. I 
> have tracked that issue down to the Pexpect upgrade in #29240 via 
> bisection. With 9.2.beta12 on the other hand, I cannot reproduce the 
> problem anymore, possibly because of the Python upgrade. There are still 
> reports about this with 9.2.beta12 though, so maybe this helps analyze the 
> problem.
>

Is singular crashing in these tests or Sage? If it's singular, it would be 
helpful to enable logging in the Pexpect interface and then trying to 
reproduce the errors without Sage


-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/bb779591-93d5-4c72-86c3-1941303a6b9ao%40googlegroups.com.

[sage-release] Re: Sage 9.2.beta12 released

[Markus Wageringel]

With a clean build on Ubuntu 20.04, ptestlong passes, but Jupyter does not 
start anymore due to an SSL problem, using Sage's Python 3. (The system 
Python 3.8 is not picked up because of missing dependencies sqlite and xz, 
but that is ok.)

$ ./sage -n jupyter
Please wait while the Sage Jupyter Notebook server starts...
Traceback (most recent call last):
  File 
"/amd/compute/mwagerin/git/sage_compute/local/lib/python3.8/site-packages/sage/repl/ipython_kernel/install.py",
 
line 307, in have_prerequisites
from notebook.notebookapp import NotebookApp
  File 
"/amd/compute/mwagerin/git/sage_compute/local/lib/python3.8/site-packages/notebook/notebookapp.py",
 
line 66, in 
from tornado import httpserver
  File 
"/amd/compute/mwagerin/git/sage_compute/local/lib/python3.8/site-packages/tornado/httpserver.py",
 
line 29, in 
import ssl
  File 
"/amd/compute/mwagerin/git/sage_compute/local/lib/python3.8/ssl.py", line 
98, in 
import _ssl # if we can't import it, let the error 
propagate
ModuleNotFoundError: No module named '_ssl'

The Jupyter notebook requires ssl, even if you do not use
https. Install the openssl development packages in your system and
then rebuild Python (sage -f python3).

Which packages need to be installed for this and why was this not a problem 
previously? OpenSSL is installed on the system:

$ openssl version
OpenSSL 1.1.1f  31 Mar 2020


On a different note, 9.2.beta11 has introduced crashes of the doctests in 
`src/sage/interfaces/singular.py` that were reproducible on about 20% of 
the test runs and have been observed by several other people as well. I 
have tracked that issue down to the Pexpect upgrade in #29240 via 
bisection. With 9.2.beta12 on the other hand, I cannot reproduce the 
problem anymore, possibly because of the Python upgrade. There are still 
reports about this with 9.2.beta12 though, so maybe this helps analyze the 
problem.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-release+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-release/ecc81f81-6fc5-4e7e-983f-7b1606e84e1an%40googlegroups.com.