On Tue, 2 Mar 2021, 03:11 brettpim, <br...@math.carleton.ca> wrote:
> I ran
>
> make toolchain
>
> and then
>
> make
>
> This was successful. I note in the log section on numpy that it is using
> a non-optimized netlib Blas.
>
OK, great.
>
> What exactly does "make toolchain" do and not do? Is there anything
> likely not to work with the sage that successfully built?
>
in particular "make toolchain" readies
pkg-config .pc files for blas, lapack, etc.
(on a modern Linux system with most recommended packages installed it does
not do much more)
numpy has a very old and a bit insane way to check for its dependencies
(basically, compilers and (c)blas/lapack)
So somehow it manages to get itself confused on your machine, when we build
it as Sage package.
Your logs say
Traceback (most recent call last):
File "../lapack_conf.py", line 7, in <module>
pc_blas = pkgconfig.parse('cblas blas')
File
"/home/brett/Projects/SAGE/sage/local/lib/python3.8/site-packages/pkgconfig/pkgconfig.py",
line 248, in parse
_raise_if_not_exists(package)
File
"/home/brett/Projects/SAGE/sage/local/lib/python3.8/site-packages/pkgconfig/pkgconfig.py",
line 103, in _raise_if_not_exists
raise PackageNotFoundError(package)
pkgconfig.pkgconfig.PackageNotFoundError: cblas not found
which indicate that said pkg-config files are not present (at least one for
cblas). They are recreated by the "make toolchain",
so that's why it all works after this. Your config.log indicates that cblas
capabilities are available in your openblas,
so it boils down to .pc file not bein set up.
The reason you see netlib Blas is due to your openblas not recognised to
have lapack capabilities,
as far as I can see from your config.log.
Maybe an Ubuntu "feature", as it's perfectly possible to have
openblas-optimised lapack included in libopenblas, maybe Ubuntu, unlike
e.g. Debian, does not do it.
Maybe it's a bug in our tests, which are admittedly a hack, introduced in
https://trac.sagemath.org/ticket/29361
This hack is something that should go once we al last deside that we should
not bother with packaging compilers
in Sage (FFS, this moment cannot be too soon, it should have happen years
ago!), as this hack is meant to cover the case of no Fortran compiler
available on the system (but I start ranting here, sorry :-)).
To test is your Sage is OK, you can run
make ptest
to run the standard Sage tests, to see if nothing is broken (takes 2-3
hours on a modern PC)
Cheers
Dima
> thanks
> brett
> On Saturday, February 27, 2021 at 8:04:18 PM UTC-6 dim...@gmail.com wrote:
>
>> this seems to be the error here:
>> https://trac.sagemath.org/ticket/31135
>>
>> I wonder if you can run
>>
>> make toolchain
>>
>> and only then run
>>
>> make
>>
>> On Sun, Feb 28, 2021 at 1:26 AM brettpim <br...@math.carleton.ca> wrote:
>> >
>> > No luck. I am not sure what to do next.
>> >
>> > On Saturday, February 27, 2021 at 6:22:35 PM UTC-6 brettpim wrote:
>> >>
>> >> True, and removing libopenblas did not help; I am still getting the
>> same error. I put libopenblas back but not numpy and am trying again
>> >>
>> >>
>> >> On Saturday, February 27, 2021 at 4:10:51 PM UTC-6 dim...@gmail.com
>> wrote:
>> >>>
>> >>> by right, one ought to read the last part of the output of
>> ./configure
>> >>> for advice on
>> >>> packages to to install, and follow it.
>> >>>
>> >>> Without libopenblas-dev installed you're going to build it, and
>> >>> everything that depends on it, from source...
>> >>>
>> >>> On Sat, Feb 27, 2021 at 9:45 PM brettpim <br...@math.carleton.ca>
>> wrote:
>> >>> >
>> >>> > I have finally tracked down and removed all installations of numpy;
>> I also removed libopenblas. I have done make bootstrap-clean distclean;
>> ./configure; and started make again. I will report the results
>> >>> >
>> >>> > On Saturday, February 27, 2021 at 2:02:59 PM UTC-6 brettpim wrote:
>> >>> >>
>> >>> >> also my git branches I am working on would be lost if I wiped
>> everything.
>> >>> >>
>> >>> >> On Saturday, February 27, 2021 at 1:50:18 PM UTC-6 brettpim wrote:
>> >>> >>>
>> >>> >>> make bootstrap-clean distclean did not solve the problem. I have
>> considered removing python3 from my system and reinstalling it but there
>> are so many other things I have that depend on python3 and would be
>> removed. I don't want to have to reinstall such a huge list. I am not sure
>> what to do next. One thought is to delete my entire sage directory and pull
>> it afresh but I am not sure this will address the issue that blas is not
>> being found properly.
>> >>> >>>
>> >>> >>>
>> >>> >>> On Saturday, February 27, 2021 at 11:18:04 AM UTC-6 brettpim
>> wrote:
>> >>> >>>>
>> >>> >>>> DIma,
>> >>> >>>>
>> >>> >>>> I found that pip installs in /usr/local when it is run as sudo.
>> I have not found anything definitive that says it should not be run as sudo
>> but just in case I sudo pip uninstalled all the packages in /usr/local and
>> pip installed them as a user which puts them all in $HOME/.local. But even
>> after this there something is still wrong with my numpy install and I am
>> still working on.
>> >>> >>>>
>> >>> >>>> While I am investigating my numpy problem I will try the serious
>> clean-up you suggest. My build was an attempt at a re-build after a git
>> pull so the clean-up is justified. I will report back
>> >>> >>>>
>> >>> >>>> thanks
>> >>> >>>> brett
>> >>> >>>> On Saturday, February 27, 2021 at 2:06:03 AM UTC-6
>> dim...@gmail.com wrote:
>> >>> >>>>>
>> >>> >>>>> On Sat, Feb 27, 2021 at 4:07 AM brettpim <
>> br...@math.carleton.ca> wrote:
>> >>> >>>>> >
>> >>> >>>>> > /usr/local/ is where pip has installed my python packages so
>> don't these files belong here?
>> >>> >>>>> >
>> >>> >>>>> >
>> >>> >>>>> > On Friday, February 26, 2021 at 5:20:03 PM UTC-6
>> dim...@gmail.com wrote:
>> >>> >>>>> >>
>> >>> >>>>> >> looks like you have a copy of (c)blas in /usr/local - and
>> this leads to trouble
>> >>> >>>>>
>> >>> >>>>> oops, sorry, it's a wrong guess, I suppose.
>> >>> >>>>>
>> >>> >>>>> What is the branch you are installing? master from github,
>> which is Sage 9.2 ?
>> >>> >>>>> Or the development branch, 9.3.beta7?
>> >>> >>>>>
>> >>> >>>>> Was it an attempt to re-build after git pull,or a fresh
>> install?
>> >>> >>>>> Fresh install should work, but for an update, you might need a
>> serious clean-up:
>> >>> >>>>>
>> >>> >>>>> make bootstrap-clean distclean
>> >>> >>>>>
>> >>> >>>>> Sage has a bit fragile system of re-creating missing .pc files
>> for
>> >>> >>>>> (c)blas/lapack, and
>> >>> >>>>> recent changes in the build system might have confused it (so
>> that's
>> >>> >>>>> why a fresh rebuild
>> >>> >>>>> might be needed)
>> >>> >>>>>
>> >>> >>>>>
>> >>> >>>>> >>
>> >>> >>>>> >> On Fri, 26 Feb 2021, 21:57 brettpim, <br...@math.carleton.ca>
>> wrote:
>> >>> >>>>> >>>
>> >>> >>>>> >>>
>> >>> >>>>> >>> I am trying to install sage from the master branch on PC
>> running Ubuntu 20.04.
>> >>> >>>>> >>> The make is failing to build numpy-1.19.1.
>> >>> >>>>> >>> It is complaining about not finding cblas or blas.
>> >>> >>>>> >>>
>> >>> >>>>> >>> In my config.log it finds openblas and so decides it will
>> not install is as SPKG
>> >>> >>>>> >>>
>> >>> >>>>> >>> but when I run python3:
>> >>> >>>>> >>> >>> import numpy
>> >>> >>>>> >>> >>> numpy.show_config()
>> >>> >>>>> >>>
>> >>> >>>>> >>> it reports
>> >>> >>>>> >>>
>> >>> >>>>> >>> blas_info:
>> >>> >>>>> >>> libraries = ['blas', 'blas']
>> >>> >>>>> >>> library_dirs = ['/usr/lib/x86_64-linux-gnu']
>> >>> >>>>> >>> include_dirs = ['/usr/local/include', '/usr/include']
>> >>> >>>>> >>> language = c
>> >>> >>>>> >>> define_macros = [('HAVE_CBLAS', None)]
>> >>> >>>>> >>> blas_opt_info:
>> >>> >>>>> >>> define_macros = [('NO_ATLAS_INFO', 1), ('HAVE_CBLAS',
>> None)]
>> >>> >>>>> >>> libraries = ['blas', 'blas']
>> >>> >>>>> >>> library_dirs = ['/usr/lib/x86_64-linux-gnu']
>> >>> >>>>> >>> include_dirs = ['/usr/local/include', '/usr/include']
>> >>> >>>>> >>> language = c
>> >>> >>>>> >>>
>> >>> >>>>> >>> which makes me think there is something wrong with blas.
>> >>> >>>>> >>>
>> >>> >>>>> >>> --
>> >>> >>>>> >>> 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/a8054ce0-69ee-4537-80ed-f51ced0f7c8cn%40googlegroups.com.
>>
>> >>> >>>>> >
>> >>> >>>>> > --
>> >>> >>>>> > 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/4c7f11e6-246d-4f58-b6d2-dfcbaf0b284bn%40googlegroups.com.
>>
>> >>> >
>> >>> > --
>> >>> > 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/249a4e39-19bb-423b-8103-1501e5c691can%40googlegroups.com.
>>
>> >
>> > --
>> > 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/9551c544-abe6-44b2-ac55-e1884d0ad060n%40googlegroups.com.
>>
>>
> --
> 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+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sage-devel/a206e982-f6b7-4d50-9334-7967c18df575n%40googlegroups.com
> <https://groups.google.com/d/msgid/sage-devel/a206e982-f6b7-4d50-9334-7967c18df575n%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
--
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+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/sage-devel/CAAWYfq1HvcGX8%2BW%2Bj41SyZh%2BkfiEBwerG3-YVPXf0TGrFC6XxA%40mail.gmail.com.
