subject:"\[sage\-devel\] Re\: #32532 \- removing gcc and gfortran spkgs"

On Friday, September 24, 2021 at 1:07:48 PM UTC-7 kcrisman wrote:

>  I guess we already have disentangled R, or else someone would have 
>> mentioned that as requiring Fortran?  
>>
>
No, we haven't even removed the R dependency yet. When I proposed 
downgrading R from its status as standard package (because of *actual* 
build problems of our R spkg at the time), it was met with opposition here 
on sage-devel. As a result, the R spkg currently has an in-between status: 
It's still a standard package but I made it "disablable" (./configure 
--disable-r). That was good enough for me as I can disable it in my 
pip-installable builds (source / wheels); and so can users who are running 
into build problems.


-- 
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/bbc6fed3-5e1b-46db-9daa-c95fa71157bcn%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

2021-09-24 Thread kcrisman

>
>
> The way to spoil users of Sage on MacOS (or anywhere) is to create a
> binary installer that work really well,
>

Correct. This is what all this discussion should be aiming toward -
including a binary installer that *allows for compiling/installing of
optional packages and Cython*. (Or gives not just a meaningful error if
you try to do that and fail, but a precise thing to click on to do it.)

The reason why building Sage from source is so important historically is
that a newbie on any platform, WITH NO EXPERIENCE USING ANY COMMAND LINE
TOOLS, was able to download something, open a Terminal, and get a version
of Sage that worked on their particular platform. And to easily update it
to the *next* version, or use Cython, or whatever.

In the event, this didn't end up being the case long-term - on Mac, at
least, largely due to Apple's doing annoying things and then partly due to
our changing how to "upgrade". But that is what one would want, to have a
"viable open-source competitor to ..."

If we can decouple those things from having to build from source, great.
It's not clear to me that is the case, because there is so much
fragmentation in the computer space. That's not our problem, but it is if
we want to attract as many users as possible - including ones who will not
be able to follow the instructions in a few posts in this thread, or who
have no choice but to stick on older "EOL" platforms - that is what is out
there. You can mock people on them, but many of them really cannot afford
to upgrade or whoever is controlling their computers will not do it.

Obviously it would good to make things less work for both our very
hard-working developers and for users. If we can disentangle from
Sage-the-distribution effectively, that is fine. Please, let's not have to
have Fortran - I guess we already have disentangled R, or else someone
would have mentioned that as requiring Fortran?

But so far, it's only been demonstrated that either 1) viable easier
distros are from projects not in the Sage core codebase like the new Mac
app that Nathan D.'s group has created (I know it wasn't just him, I just
forget who else) or like Isuru's distribution 2) you need to be well within
the Linux ecosystem and conversant with things that a large number of users
will not be familiar with. Including some who are highly
computer-literate, but not shell-literate. And having the gcc/gfortran,
again historically, was a great way to fix that. Apparently that is still
the case in a number of situations, or we wouldn't be having this
conversation. It's a failsafe for those times, which keep creeping in, and
unless Apple itself starts providing the full ecosystem (fat chance), we'll
still be relying on the good graces of projects like Homebrew.

The point is made several times that an internet solution may work for
this. But I have to point out YET AGAIN that not everyone lives in happy
internet-land with great connectivity all day long. And Cocalc is a big
app for some connectivity situations too ... so that's not the long-term
answer. As many of us discovered during the pandemic and all of a sudden
even some colleagues in "the West" (whatever that means) did not have as
good of internet as we thought.

--
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/c2bfb4c7-fa05-4b59-9c2a-7a5755bb22dfn%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Friday, September 24, 2021 at 12:28:02 PM UTC-7 Matthias Koeppe wrote:

> On Friday, September 24, 2021 at 11:12:38 AM UTC-7 wst...@gmail.com wrote:
>
>> I also personally wish there were a significantly smaller Sage core with 
>> much less 
>> dependencies, and which removes everything from the Sage that annoys 
>> Dima, and much more.This is a difficult
>> technical challenge, since it would certainly involve changing core parts 
>> of the library.  E.g., it would be nice to have
>> a working Sage that doesn't depend on Maxima or GAP being present, but 
>> still starts up and is generally useful
>> for the rest of what Sage does.Creating such a thing involves making 
>> significant changes to the assumptions that
>> can be made in the Sage library code about what they assume is available 
>> by default. 
>>
>
> Yes, this is exactly what modularization is doing. 
> https://trac.sagemath.org/ticket/29705
>

It works by defining subsets (namespace packages) of the Sage library that 
are carefully designed around the dependencies on compiled 
(non-pip-installable) libraries.

For example, https://trac.sagemath.org/ticket/32432 defines a distribution 
sagemath-polyhedra with minimal dependencies. This work needs help in 
disentangling some basic library code (such as for integers, integer 
matrices, etc.) that depends on too many libraries simultaneously (PARI and 
FLINT and NTL and )

The benefit of constructing such distributions is: It has a clearly 
definable goal; when done, it can immediately be advertised to potential 
new users; and it does not compromise the usefulness of the Sage 
distribution for old users on the way there.

-- 
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/5fa26a2e-5987-4f03-be2a-ef729d326bb2n%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Fri, 24 Sep 2021, 19:12 William Stein,  wrote:

>
>
> On Fri, Sep 24, 2021 at 10:38 AM Dima Pasechnik  wrote:
>
>>
>>
>> On Fri, Sep 24, 2021 at 6:26 PM William Stein  wrote:
>>
>>>
>>>
>>> On Fri, Sep 24, 2021 at 9:03 AM Dima Pasechnik  wrote
>>>
>>
>
>> yes, Sage binaries on macOS in particular are total crap - in part thanks
>> to "let's vendor everything moving" approach,
>> in part thanks to Apple being bad to developers.
>>
>
> I assume you are talking about the official binaries that are distributed
> on Sagemath.org.  Fortunately, the Sage binaries on
> MacOS that are produced by the conda-forge devs are not total crap.   If
> you read
>
>
> https://groups.google.com/g/sage-devel/c/QeYle_D8Otc/m/_5Q8zHPLAwAJ?utm_medium=email_source=footer
>
> you'll see that Isuru Fernando quickly responded to my post with something
> I could copy/paste into the terminal on my Macbook
> M1 ARM laptop.  I did so, waited for the download/install, and had what
> appeared to be a perfectly working self-contained pure userspace
> copy of Sage on my M1 laptop (and it is not using x86_64 emulation).
> Similarly, the "JupyterLab app" that sparked that thread
> is also a nice graphical MacOS installer for JupyterLab (via Electron)
> with the Python scientific stack bundled -- the guy
> who made it was introducing it in the JupyterLab weekly dev meeting, and
> within a few sentences of him telling us about it,
> I had installed and tested it and it  was working.  He also said that it
> wasn't particular hard for him to build, since so much work
> has gone into making things like conda-forge more mature.
>
>
>>
>>
>>
>>> Hopefully going forward, we can spoil users of Sage even more than we
>>> already are.
>>>
>>
>> Sage is trying to bite more than it can chew, and it is badly needs to
>> slim down, before becoming a total crap nobody uses.
>> If one wants to spoil (in the good way) Sage users on macOS, more brew
>> formulas are needed for Sage packages.
>> Then a similar to Sage on Debian Sage package might be possible.
>> This is totally orthogonal to vendoring things needlessly.
>>
>
> The way to spoil users of Sage on MacOS (or anywhere) is to create a
> binary installer that work really well, similar to
> what the JupyterLab devs did in [3].  It seems to me that this doesn't
> necessarily involve Homebrew or Debian Sage at all, but
> instead involve vendoring things, though that vendoring might better
> happen by joining forces with conda-forge, rather than
> duplicating their work.
>
> Isn't conda-forge also doing exactly the thing that you're arguing against
> as doing?
>

Yes, they surely are, and I am saying that we should not duplicate their
(or other distros') efforts, and join forces instead.

The sooner dead weight packages such as gcc are removed from Sage, the
better, not the least because they allow to keep dragging our feet...


I started Sage and its package system around the same time Travis Oliphant
> started conda, and it
> could have turned out that Sage's package system is the one that has
> 14000+ packages and everybody uses in the scientific
> Python ecosystem.   It didn't turn out that way.  I'm glad some people
> such as Isuru in the Sage community have
> embraced conda-forge.
>
> I hope we can focus on solving the big important problems, rather than
> whatever is being argued about in this thread.
>
> Matthias is right that another important big challenge is "pip install
> sagemath", and it sounds to me like he has a viable
>
>
>
>
>
> strategy to eventually solve that problem.  It's a completely different
> problem than "binary that is easy for Ken Ribet
> to install on macOS", which conda-forge could solve.   And yet a different
> big problem is a "Sage on Microsoft Windows",
> which conda-forge doesn't solve at all, and for which the cygwin solution
> isn't close to optimal.
>
> In conflict with all of the above, I also personally wish there were a
> significantly smaller Sage core with much less
> dependencies, and which removes everything from the Sage that annoys Dima,
> and much more.This is a difficult
> technical challenge, since it would certainly involve changing core parts
> of the library.  E.g., it would be nice to have
> a working Sage that doesn't depend on Maxima or GAP being present, but
> still starts up and is generally useful
> for the rest of what Sage does.Creating such a thing involves making
> significant changes to the assumptions that
> can be made in the Sage library code about what they assume is available
> by default.
>
>  -- William
>
>
>>
>>> [1] A good example of a representative influential mathematician, e.g.,
>>> he was the last president of the American Mathematical Society.  He uses
>>> MacOS.
>>> [2] https://github.com/jupyterlab/jupyterlab_app
>>> [3]
>>> https://groups.google.com/g/sage-devel/c/QeYle_D8Otc/m/_5Q8zHPLAwAJ?utm_medium=email_source=footer
>>>
>>> --
>>> William (http://wstein.org)
>>>
>>> --
>>> You received this message because you

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Friday, September 24, 2021 at 11:12:38 AM UTC-7 wst...@gmail.com wrote:

> I also personally wish there were a significantly smaller Sage core with 
> much less 
> dependencies, and which removes everything from the Sage that annoys Dima, 
> and much more.This is a difficult
> technical challenge, since it would certainly involve changing core parts 
> of the library.  E.g., it would be nice to have
> a working Sage that doesn't depend on Maxima or GAP being present, but 
> still starts up and is generally useful
> for the rest of what Sage does.Creating such a thing involves making 
> significant changes to the assumptions that
> can be made in the Sage library code about what they assume is available 
> by default. 
>

Yes, this is exactly what modularization is 
doing. https://trac.sagemath.org/ticket/29705



-- 
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/3ed786da-d0ea-4938-a60e-8820e6f8f561n%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Fri, 24 Sep 2021, 18:26 William Stein,  wrote:

>
>
> On Fri, Sep 24, 2021 at 9:03 AM Dima Pasechnik  wrote
> > We spoiled the users, they assumed that Sage is a distribution. Yes,
> there will be a short while they'll ask - and the answer will be:
> > RTFM [...]
>
> I don't want to do that to our users.  Sage was started to compete with
> [Magma | Mathematica | Maple], so the install process for Sage needed to
> compete with what [Magma | Mathematica | Maple] offers.
>

we already have https://trac.sagemath.org/ticket/32060
which will be, from the point of view of 99.9% of the users, identical to
what's proposed in the ticket we are discussing, #32532.
(the remaining 0.1% are people deliberately building gcc/gfortran as a part
of Sage).

Thus on Matthias' side it's a bit strange to claim that with #32532 we will
be hit by a barrage of users' questions. We will be, as much as, or as
little as, we will be with #32060.


 A few weeks ago, I was bummed that Ken Ribet [1] was lamenting on Facebook
> that he does NOT recommend that people install Sage,
>

by the way, that's a very kind feedback indeed, let's just all lament on
social media about entitled Sage users who contribute laments to Sage. :-)

 Maybe they can at least contribute to Sage donation campaign some money
for developers' Apple fees and a few Apple M1 macs...


because it is too confusing or likely to be broken.  Something like [2, 3]
> but for Sage would go a long way to help.
>
> Hopefully going forward, we can spoil users of Sage even more than we
> already are.
>
> [1] A good example of a representative influential mathematician, e.g., he
> was the last president of the American Mathematical Society.  He uses MacOS.
> [2] https://github.com/jupyterlab/jupyterlab_app
> [3]
> https://groups.google.com/g/sage-devel/c/QeYle_D8Otc/m/_5Q8zHPLAwAJ?utm_medium=email_source=footer
>
> --
> William (http://wstein.org)
>
> --
> 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/CACLE5GB1PuUXJcU-d-PrOsYo1yL3BTeaJPsVsSq-3nGzT9095Q%40mail.gmail.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/CAAWYfq0gbAW0T9cPnpHS%3DEffkYLnjX6RaBPo45u_%2BZYYg8X-Zw%40mail.gmail.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

2021-09-24 Thread Nils Bruin

On Friday, 24 September 2021 at 11:12:38 UTC-7 wst...@gmail.com wrote:

>
> In conflict with all of the above, I also personally wish there were a 
> significantly smaller Sage core with much less 
> dependencies, and which removes everything from the Sage that annoys Dima, 
> and much more.This is a difficult
> technical challenge, since it would certainly involve changing core parts 
> of the library.  E.g., it would be nice to have
> a working Sage that doesn't depend on Maxima or GAP being present, but 
> still starts up and is generally useful
> for the rest of what Sage does.Creating such a thing involves making 
> significant changes to the assumptions that
> can be made in the Sage library code about what they assume is available 
> by default. 
>
> Be careful what you wish for! Although in this particular case you may not 
get it anyway, so it wouldn't be a problem. If there are different degrees 
to which you can have "sage installed" there's a whole new  slew of 
problems that arise: if someone has a problem in such a situation, they 
need to figure if they have a partial sagemath install and if so, which 
parts are missing and if their absence causes the problem. I think this 
would similarly lead to people not recommending to install sagemath. I 
think an inevitable corollary of having components of sage not mandatory is 
that support and maintenance of these components deteriorates.

Note that Ken not advising people to install sage doesn't necessarily mean 
he doesn't advise people to use it. For many people who are not 
particularly computer-literate, using cloud-based solutions nowadays may be 
a sensible option. I think for sage there is something called CoCalc that 
might do a reasonable job.

I do think it is important that we do package sage in a way that  a lot of 
people can productively install it, though, and especially that it's easy 
to get a development-capable installation. On a positive note, a student of 
mine was able to get  a source install on Windows via cygwin running 
recently. I understand it wasn't pleasant experience, but the student did 
succeed without outside help -- I was impressed. I would have advised WSL2, 
which worked quite well for me when I tried it on Win10. 

He did note that the Cygwin binary distributions available at 
https://github.com/sagemath/sage-windows/releases only go to 9.3 -- and 
there were some functional differences between 9.3 and 9.4 that lead to 
incompatibilities for https://github.com/nbruin/RiemannTheta that should 
eventually get incorporated in sage to resolve the 12-year old ticket 
https://trac.sagemath.org/ticket/6371. This is partially an experiment to 
see if offering new functionality initially as a separate pip-installable 
package improves development and early availability [like a preprint server 
...]. One big downside I already found is that "sage -pip install ... 
--user" doesn't work. People need a LOT of knowledge to use a sagemath 
python module that is not installed in the main sage tree (which may not be 
writable for them!).

-- 
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/01cf2804-acbe-4c7d-939a-b4fef53d610an%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

2021-09-24 Thread William Stein

On Fri, Sep 24, 2021 at 10:38 AM Dima Pasechnik  wrote:

>
>
> On Fri, Sep 24, 2021 at 6:26 PM William Stein  wrote:
>
>>
>>
>> On Fri, Sep 24, 2021 at 9:03 AM Dima Pasechnik  wrote
>>
>

> yes, Sage binaries on macOS in particular are total crap - in part thanks
> to "let's vendor everything moving" approach,
> in part thanks to Apple being bad to developers.
>

I assume you are talking about the official binaries that are distributed
on Sagemath.org.  Fortunately, the Sage binaries on
MacOS that are produced by the conda-forge devs are not total crap.   If
you read

https://groups.google.com/g/sage-devel/c/QeYle_D8Otc/m/_5Q8zHPLAwAJ?utm_medium=email_source=footer

you'll see that Isuru Fernando quickly responded to my post with something
I could copy/paste into the terminal on my Macbook
M1 ARM laptop.  I did so, waited for the download/install, and had what
appeared to be a perfectly working self-contained pure userspace
copy of Sage on my M1 laptop (and it is not using x86_64 emulation).
Similarly, the "JupyterLab app" that sparked that thread
is also a nice graphical MacOS installer for JupyterLab (via Electron) with
the Python scientific stack bundled -- the guy
who made it was introducing it in the JupyterLab weekly dev meeting, and
within a few sentences of him telling us about it,
I had installed and tested it and it  was working.  He also said that it
wasn't particular hard for him to build, since so much work
has gone into making things like conda-forge more mature.

>
>
>
>> Hopefully going forward, we can spoil users of Sage even more than we
>> already are.
>>
>
> Sage is trying to bite more than it can chew, and it is badly needs to
> slim down, before becoming a total crap nobody uses.
> If one wants to spoil (in the good way) Sage users on macOS, more brew
> formulas are needed for Sage packages.
> Then a similar to Sage on Debian Sage package might be possible.
> This is totally orthogonal to vendoring things needlessly.
>

The way to spoil users of Sage on MacOS (or anywhere) is to create a binary
installer that work really well, similar to
what the JupyterLab devs did in [3].  It seems to me that this doesn't
necessarily involve Homebrew or Debian Sage at all, but
instead involve vendoring things, though that vendoring might better happen
by joining forces with conda-forge, rather than
duplicating their work.

Isn't conda-forge also doing exactly the thing that you're arguing against
as doing?
I started Sage and its package system around the same time Travis Oliphant
started conda, and it
could have turned out that Sage's package system is the one that has 14000+
packages and everybody uses in the scientific
Python ecosystem.   It didn't turn out that way.  I'm glad some people such
as Isuru in the Sage community have
embraced conda-forge.

I hope we can focus on solving the big important problems, rather than
whatever is being argued about in this thread.

Matthias is right that another important big challenge is "pip install
sagemath", and it sounds to me like he has a viable

strategy to eventually solve that problem.  It's a completely different
problem than "binary that is easy for Ken Ribet
to install on macOS", which conda-forge could solve.   And yet a different
big problem is a "Sage on Microsoft Windows",
which conda-forge doesn't solve at all, and for which the cygwin solution
isn't close to optimal.

In conflict with all of the above, I also personally wish there were a
significantly smaller Sage core with much less
dependencies, and which removes everything from the Sage that annoys Dima,
and much more.This is a difficult
technical challenge, since it would certainly involve changing core parts
of the library.  E.g., it would be nice to have
a working Sage that doesn't depend on Maxima or GAP being present, but
still starts up and is generally useful
for the rest of what Sage does.Creating such a thing involves making
significant changes to the assumptions that
can be made in the Sage library code about what they assume is available by
default.

 -- William

>
>> [1] A good example of a representative influential mathematician, e.g.,
>> he was the last president of the American Mathematical Society.  He uses
>> MacOS.
>> [2] https://github.com/jupyterlab/jupyterlab_app
>> [3]
>> https://groups.google.com/g/sage-devel/c/QeYle_D8Otc/m/_5Q8zHPLAwAJ?utm_medium=email_source=footer
>>
>> --
>> William (http://wstein.org)
>>
>> --
>> 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/CACLE5GB1PuUXJcU-d-PrOsYo1yL3BTeaJPsVsSq-3nGzT9095Q%40mail.gmail.com
>> 
>>

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Fri, Sep 24, 2021 at 6:26 PM William Stein  wrote:

>
>
> On Fri, Sep 24, 2021 at 9:03 AM Dima Pasechnik  wrote
> > We spoiled the users, they assumed that Sage is a distribution. Yes,
> there will be a short while they'll ask - and the answer will be:
> > RTFM [...]
>
> I don't want to do that to our users.  Sage was started to compete with
> [Magma | Mathematica | Maple], so the install process for Sage needed to
> compete with what [Magma | Mathematica | Maple] offers.   A few weeks ago,
> I was bummed that Ken Ribet [1] was lamenting on Facebook that he does NOT
> recommend that people install Sage, because it is too confusing or likely
> to be broken.  Something like [2, 3] but for Sage would go a long way to
> help.
>

Does Ken build Sage from source? I guess he does not, and
yes, Sage binaries on macOS in particular are total crap - in part thanks
to "let's vendor everything moving" approach,
in part thanks to Apple being bad to developers.



> Hopefully going forward, we can spoil users of Sage even more than we
> already are.
>

Sage is trying to bite more than it can chew, and it is badly needs to slim
down, before becoming a total crap nobody uses.
If one wants to spoil (in the good way) Sage users on macOS, more brew
formulas are needed for Sage packages.
Then a similar to Sage on Debian Sage package might be possible.
This is totally orthogonal to vendoring things needlessly.


> [1] A good example of a representative influential mathematician, e.g., he
> was the last president of the American Mathematical Society.  He uses MacOS.
> [2] https://github.com/jupyterlab/jupyterlab_app
> [3]
> https://groups.google.com/g/sage-devel/c/QeYle_D8Otc/m/_5Q8zHPLAwAJ?utm_medium=email_source=footer
>
> --
> William (http://wstein.org)
>
> --
> 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/CACLE5GB1PuUXJcU-d-PrOsYo1yL3BTeaJPsVsSq-3nGzT9095Q%40mail.gmail.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/CAAWYfq0hT_e-2Mq1B4MVqEq1P5a929HHUo0bq3z-%2BexK36diRA%40mail.gmail.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

2021-09-24 Thread William Stein

On Fri, Sep 24, 2021 at 9:03 AM Dima Pasechnik  wrote
> We spoiled the users, they assumed that Sage is a distribution. Yes,
there will be a short while they'll ask - and the answer will be:
> RTFM [...]

I don't want to do that to our users.  Sage was started to compete with
[Magma | Mathematica | Maple], so the install process for Sage needed to
compete with what [Magma | Mathematica | Maple] offers.   A few weeks ago,
I was bummed that Ken Ribet [1] was lamenting on Facebook that he does NOT
recommend that people install Sage, because it is too confusing or likely
to be broken.  Something like [2, 3] but for Sage would go a long way to
help.

Hopefully going forward, we can spoil users of Sage even more than we
already are.

[1] A good example of a representative influential mathematician, e.g., he
was the last president of the American Mathematical Society.  He uses MacOS.
[2] https://github.com/jupyterlab/jupyterlab_app
[3]
https://groups.google.com/g/sage-devel/c/QeYle_D8Otc/m/_5Q8zHPLAwAJ?utm_medium=email_source=footer

-- 
William (http://wstein.org)

-- 
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/CACLE5GB1PuUXJcU-d-PrOsYo1yL3BTeaJPsVsSq-3nGzT9095Q%40mail.gmail.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Fri, Sep 24, 2021 at 5:48 PM Matthias Koeppe 
wrote:

> On Friday, September 24, 2021 at 9:36:57 AM UTC-7 Dima Pasechnik wrote:
>
>> On Fri, 24 Sep 2021, 17:30 Matthias Koeppe,  wrote:
>>
>>> You might recall that in 2020 I wrote the system-package recommendation
>>> scripts
>>>
>>
>>
> Your recommendations get routinely ignored, as they are recommendations,
>> and not laws - one can proceed to building, and ignore them. On the other
>> hand, if you try to build Sage without any C compiler on the PATH, you'll
>> be forced to install one, in no uncertain terms.
>>
>
> Yes, as you know I implemented this already in
> https://trac.sagemath.org/ticket/32060 ("configure: Change defaults to
> --with-system-gcc=force"), which is slated for inclusion in Sage 9.5.
> There is no need to put the axe to the Sage distribution just because we
> want issue more forceful recommendations.
>

with a number of platforms needing a bump to gcc 11 (e.g. Apple's M1), it's
better to stop fooling around with vendoring gcc all together; the earlier,
the better.



>
>
> --
> 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/24a4706e-aa55-4dc4-aab8-2d1276a20e04n%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/CAAWYfq1P8OQ8bQPQ2UxOv2skdR4%2BC_z-VqANJ8KqKvTxUdtoig%40mail.gmail.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Friday, September 24, 2021 at 9:36:57 AM UTC-7 Dima Pasechnik wrote:

> On Fri, 24 Sep 2021, 17:30 Matthias Koeppe,  wrote:
>
>> You might recall that in 2020 I wrote the system-package recommendation 
>> scripts
>>
>  
>
Your recommendations get routinely ignored, as they are recommendations, 
> and not laws - one can proceed to building, and ignore them. On the other 
> hand, if you try to build Sage without any C compiler on the PATH, you'll 
> be forced to install one, in no uncertain terms.
>

Yes, as you know I implemented this already 
in https://trac.sagemath.org/ticket/32060 ("configure: Change defaults to 
--with-system-gcc=force"), which is slated for inclusion in Sage 9.5.
There is no need to put the axe to the Sage distribution just because we 
want issue more forceful recommendations.


-- 
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/24a4706e-aa55-4dc4-aab8-2d1276a20e04n%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Fri, 24 Sep 2021, 17:30 Matthias Koeppe, 
wrote:

> On Friday, September 24, 2021 at 9:03:20 AM UTC-7 Dima Pasechnik wrote:
>
>> On Fri, Sep 24, 2021 at 4:43 PM Matthias Koeppe 
>> wrote:
>>
>>> you are shifting the work from our working spkg-install.in script to
>>> handholding users who attempt [install of gcc from source] on the mailing
>>> lists.
>>>
>>
>> We spoiled the users, they assumed that Sage is a distribution. Yes,
>> there will be a short while they'll ask - and the answer will be:
>> RTFM/apt-get install gfortran/brew install gcc/...
>>
>
> You might recall that in 2020 I wrote the system-package recommendation
> scripts, a virtual beekeeper if you will, to reduce the workload of those
> answering the repeated questions by users on the mailing list regarding
> installation problems.
>

My surname means "beekeeper" in Russian/Ukrainian, FYI :-P

Your recommendations get routinely ignored, as they are recommendations,
and not laws - one can proceed to building, and ignore them. On the other
hand, if you try to build Sage without any C compiler on the PATH, you'll
be forced to install one, in no uncertain terms.




It would be unwise to make changes to the Sage distribution that provoke a
> new wave of installation problems.
>
> --
> 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/6ab95d5b-6e54-4e2a-aaf8-19e6f858c9a9n%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/CAAWYfq2zcbFkELCu0KgyhS1F%2BOoi4qgKOBJQhMvLgYomY0qQWA%40mail.gmail.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Friday, September 24, 2021 at 9:03:20 AM UTC-7 Dima Pasechnik wrote:

> On Fri, Sep 24, 2021 at 4:43 PM Matthias Koeppe  
> wrote:
>
>> you are shifting the work from our working spkg-install.in script to 
>> handholding users who attempt [install of gcc from source] on the mailing 
>> lists.
>>
>
> We spoiled the users, they assumed that Sage is a distribution. Yes, there 
> will be a short while they'll ask - and the answer will be: 
> RTFM/apt-get install gfortran/brew install gcc/...
>

You might recall that in 2020 I wrote the system-package recommendation 
scripts, a virtual beekeeper if you will, to reduce the workload of those 
answering the repeated questions by users on the mailing list regarding 
installation problems.
It would be unwise to make changes to the Sage distribution that provoke a 
new wave of installation problems.

-- 
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/6ab95d5b-6e54-4e2a-aaf8-19e6f858c9a9n%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Fri, Sep 24, 2021 at 4:50 PM Matthias Koeppe 
wrote:

> On Thursday, September 23, 2021 at 9:09:45 PM UTC-7 Michael Orlitzky wrote:
>
>> [Matthias'] Item #2 happened, incidentally, only because we've copy &
>> pasted so
>> many packages into sage pretending to be a linux distribution.
>> Eliminating that problem altogether begins right here, at the root of
>> the dependency tree.
>
>
> In another post:
>
> > [Matthias's items] 2, 4, 6, and 7 are addressed by Conda, Nix, Guix,
> Homebrew, or even
> > Gentoo Prefix. These projects *want* to build entire distributions that
> > run out of your home directory.
>
> These two comments together really are the crux of the discussion.
>
> If we say that availability of gfortran in any of these user-installable
> distributions is the solution, then this applies to lots of other spkgs as
> well --- perhaps to ALL of our non-Python packages.
>
> So essentially it is to say, let's stop maintaining the Sage distribution.
>

Sage "distribution" is what's in build/ (and in upstream/). Many packages
don't need to be vendored, like we stopped vendoring tar, and we can stop
vendoring patch, gcc, gfortran, python3, and few other packages, very
easily.

Of course these scripts still need to be maintained, so that's not "let's
stop maintaining", it's "let's stop vendoring".




>
> This certainly *could* make sense for the project -- but a lot of work is
> needed to bring missing packages to these distributions: conda-forge does
> not have all of our optional packages, homebrew is missing a lot of
> packages. On the Sage side, also working on the tasks
> https://trac.sagemath.org/ticket/30914 (creating proper upstreams for
> some Sage-specific packages) will help.
>
> The big problem is that the middle ground between "Complete Sage
> distribution that works in most use cases" and "No Sage distribution" is
> worse than both of the extremes. By removing spkgs one by one, we would
> make the Sage distribution less useful.
> So this is not a meaningful process.
>

>
>
> --
> 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/323c5a9d-cdd4-42ee-b250-98d2c5caea4bn%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/CAAWYfq0E%3DyYqyMpj4%2B2pzvGzu4sW59z37rNkoqFpB_hp4qnhnA%40mail.gmail.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thursday, September 23, 2021 at 9:13:23 PM UTC-7 Michael Orlitzky wrote:

> > > Whoever is building the binary images could install a compiler into 
> > > SAGE_LOCAL with [some system package command] or whatever before they 
> start. There's no 
> > > need for it to come from an SPKG.
>

Well, we have several binary distributors now: (1) Whoever is making the 
binaries for Linux (I presume it is Volker); (2) the new macOS distribution 
by Marc Culler (https://github.com/3-manifolds/Sage_macOS/releases); (3) 
the relocatable wheels from my ticket 
https://trac.sagemath.org/ticket/31396 
Making gfortran each of these distributors' individual problems is a step 
in the wrong direction. 

-- 
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/2cb8287a-0c82-4632-9aec-2b9e1665983fn%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Friday, September 24, 2021 at 9:03:20 AM UTC-7 Dima Pasechnik wrote:

> NumPy/SciPy mailing lists/support forums aren't full of questions on how 
> to install a Fortran compiler, you know.
>

Yes because of wheels.
 

-- 
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/5d49222b-8b03-4f5f-8b2d-7a79e431638cn%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Fri, Sep 24, 2021 at 4:43 PM Matthias Koeppe 
wrote:

> On Friday, September 24, 2021 at 8:27:07 AM UTC-7 Dima Pasechnik wrote:
>
>> On Fri, Sep 24, 2021 at 4:14 PM Matthias Koeppe 
>> wrote:
>>
>>> Bizarre that you think that it's easy for users to build gcc from source
>>> but it's somehow hard for us to maintain the package that builds gcc from
>>> source.
>>>
>>> Maintaining is a task. Non-maintaining is no task to worry about.
>> What's bizarre about this?
>>
>
> Because you are shifting the work from our working spkg-install.in script
> to handholding users who attempt it on the mailing lists.
>

We spoiled the users, they assumed that Sage is a distribution. Yes, there
will be a short while they'll ask - and the answer will be:
RTFM/apt-get install gfortran/brew install gcc/...

NumPy/SciPy mailing lists/support forums aren't full of questions on how to
install a Fortran compiler, you know.


>
> --
> 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/22649531-58ee-4cc8-a717-79c1d986b50an%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/CAAWYfq1dOKMsgBuCddyJKtx%2Bh%3D-QvM0UJxPfiyc7VhfUAhqLCQ%40mail.gmail.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thursday, September 23, 2021 at 9:09:45 PM UTC-7 Michael Orlitzky wrote:

> [Matthias'] Item #2 happened, incidentally, only because we've copy & 
> pasted so 
> many packages into sage pretending to be a linux distribution. 
> Eliminating that problem altogether begins right here, at the root of 
> the dependency tree.

In another post:

> [Matthias's items] 2, 4, 6, and 7 are addressed by Conda, Nix, Guix, 
Homebrew, or even
> Gentoo Prefix. These projects *want* to build entire distributions that
> run out of your home directory.

These two comments together really are the crux of the discussion.

If we say that availability of gfortran in any of these user-installable 
distributions is the solution, then this applies to lots of other spkgs as 
well --- perhaps to ALL of our non-Python packages.

So essentially it is to say, let's stop maintaining the Sage distribution. 

This certainly *could* make sense for the project -- but a lot of work is 
needed to bring missing packages to these distributions: conda-forge does 
not have all of our optional packages, homebrew is missing a lot of 
packages. On the Sage side, also working on the tasks 
https://trac.sagemath.org/ticket/30914 (creating proper upstreams for some 
Sage-specific packages) will help.

The big problem is that the middle ground between "Complete Sage 
distribution that works in most use cases" and "No Sage distribution" is 
worse than both of the extremes. By removing spkgs one by one, we would 
make the Sage distribution less useful. 
So this is not a meaningful process.

-- 
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/323c5a9d-cdd4-42ee-b250-98d2c5caea4bn%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Friday, September 24, 2021 at 8:27:07 AM UTC-7 Dima Pasechnik wrote:

> On Fri, Sep 24, 2021 at 4:14 PM Matthias Koeppe  
> wrote:
>
>> Bizarre that you think that it's easy for users to build gcc from source 
>> but it's somehow hard for us to maintain the package that builds gcc from 
>> source.
>>
>> Maintaining is a task. Non-maintaining is no task to worry about. 
> What's bizarre about this? 
>

Because you are shifting the work from our working spkg-install.in script 
to handholding users who attempt it on the mailing lists.

-- 
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/22649531-58ee-4cc8-a717-79c1d986b50an%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Friday, September 24, 2021 at 8:30:40 AM UTC-7 Dima Pasechnik wrote:

> On Fri, Sep 24, 2021 at 3:54 PM Matthias Koeppe  
> wrote:
>
>> On Friday, September 24, 2021 at 4:23:32 AM UTC-7 Dima Pasechnik wrote:
>>
>>> Why does it build everything from source? E.g. numpy doesn't build 
>>> Fortran compilers.
>>> Why do we have to outdo other Python projects here?
>>>
>>
>> I have explained in the paragraph that you quoted (and previously in 
>> #32532). Wheels. They have wheels. We don't.
>>
>
> Are we supposed to reinvent wheels too (pun intended)?
>

Fortunately no -- Python wheels are sufficiently flexible. We would "just" 
need a strategy for deployment of Sage components as wheels, and people who 
work on it. There are many tasks and opportunities. Meta-ticket 
https://trac.sagemath.org/ticket/31251 is an entry point.
 

> You never explained why you went with building everything from source, and 
> why you think it's a good idea (I don't think it is).
> Do explain why we cannot have what they have.
>

Oh, wheel deployment of Sage-the-distribution is waiting for review 
at https://trac.sagemath.org/ticket/31396



-- 
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/782b8775-8a58-4820-96c4-9d91aa4d608bn%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Fri, Sep 24, 2021 at 3:54 PM Matthias Koeppe 
wrote:

> On Friday, September 24, 2021 at 4:23:32 AM UTC-7 Dima Pasechnik wrote:
>
>>
>> On Fri, 24 Sep 2021, 02:12 Matthias Koeppe,  wrote:
>>
>>> I repeat my strong objections to the proposed change -- which does not
>>> solve any problems and only creates new ones.
>>> 7. (As I explained before in
>>> https://trac.sagemath.org/ticket/32532#comment:11) in particular,
>>> Python users / developers generally have a working gcc (because many Python
>>> packages include extension modules that need to be compiled) but they have
>>> no need to have gfortran installed. The reason is that the few Python
>>> packages that use gfortran (such as scipy) have high-quality deployments of
>>> wheels to PyPI, which are widely used. Sage, however, does not yet have a
>>> working strategy that would enable us to use wheels from PyPI. As you know,
>>> I have been working on a PyPI-based deployment solution for Sage in
>>> https://trac.sagemath.org/ticket/29039; this builds everything from
>>> source.
>>>
>>
>> Why does it build everything from source? E.g. numpy doesn't build
>> Fortran compilers.
>> Why do we have to outdo other Python projects here?
>>
>
> I have explained in the paragraph that you quoted (and previously in
> #32532). Wheels. They have wheels. We don't.
>

Are we supposed to reinvent wheels too (pun intended)?

You never explained why you went with building everything from source, and
why you think it's a good idea (I don't think it is).
Do explain why we cannot have what they have.



>
>
>
>
> --
> 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/bffb95ab-3407-487f-8e19-b8945f7e10c3n%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/CAAWYfq1L5oSJimRYqXp_5%3D4X8RNNv7n%2Bf2am%3DdihzRb_eMN00w%40mail.gmail.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Fri, Sep 24, 2021 at 4:14 PM Matthias Koeppe 
wrote:

> On Friday, September 24, 2021 at 4:23:32 AM UTC-7 Dima Pasechnik wrote:
>
>>
>> On Fri, 24 Sep 2021, 02:12 Matthias Koeppe,  wrote:
>>
>>> I repeat my strong objections to the proposed change -- which does not
>>> solve any problems and only creates new ones.
>>>
>>> 2. We used the gcc spkg as recently as this Spring, for the 9.3 release,
>>> in order to support Fedora 34, which had just switched to gcc 11. No other
>>> solution was proposed or feasible at the time because several of our
>>> standard packages were not ready for this compiler, and a Sage release was
>>> already long overdue.
>>>
>>
>> nobody pushes us into not skipping one particular version.
>> I cannot imagine e.g. cpython or scipy delaying a release in such a
>> situation.
>> Besides, building gcc from source is not too hard
>> with a modern C compiler available.
>>
>
> Bizarre that you think that it's easy for users to build gcc from source
> but it's somehow hard for us to maintain the package that builds gcc from
> source.
>
> Maintaining is a task. Non-maintaining is no task to worry about.
What's bizarre about this?


>
> --
> 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/01d0a820-7967-45e8-bd72-f75e3782ebaen%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/CAAWYfq0t3%3DWNU_E4ryk0BJvRdqa7qca0CuFZdLzsnSGSmpVz7w%40mail.gmail.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Friday, September 24, 2021 at 4:23:32 AM UTC-7 Dima Pasechnik wrote:

>
> On Fri, 24 Sep 2021, 02:12 Matthias Koeppe,  wrote:
>
>> I repeat my strong objections to the proposed change -- which does not 
>> solve any problems and only creates new ones.
>>
>> 2. We used the gcc spkg as recently as this Spring, for the 9.3 release, 
>> in order to support Fedora 34, which had just switched to gcc 11. No other 
>> solution was proposed or feasible at the time because several of our 
>> standard packages were not ready for this compiler, and a Sage release was 
>> already long overdue.
>>
>
> nobody pushes us into not skipping one particular version. 
> I cannot imagine e.g. cpython or scipy delaying a release in such a 
> situation.
> Besides, building gcc from source is not too hard
> with a modern C compiler available.
>

Bizarre that you think that it's easy for users to build gcc from source 
but it's somehow hard for us to maintain the package that builds gcc from 
source.


-- 
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/01d0a820-7967-45e8-bd72-f75e3782ebaen%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Friday, September 24, 2021 at 6:28:50 AM UTC-7 Julien Puydt wrote:

> I made the point years ago that sagemath should be broken in two: 
>
> - sagemath-the-software ; 
>
> - sagemath-the-distribution. 
>
> This thread is an illustration of why that would simplify matters
>

Yes, and many of us have worked on making progress in this direction.

But the approach of chipping away parts of the Sage distribution, declaring 
that this or that package is one everyone's system, does not help in any 
way with this.

Meta-ticket https://trac.sagemath.org/ticket/29705 lists many tasks that 
would actually help.

-- 
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/a3ea275b-6fe3-4e3d-8109-a5fd09def812n%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Friday, September 24, 2021 at 4:23:32 AM UTC-7 Dima Pasechnik wrote:

>
> On Fri, 24 Sep 2021, 02:12 Matthias Koeppe,  wrote:
>
>> I repeat my strong objections to the proposed change -- which does not 
>> solve any problems and only creates new ones.
>> 7. (As I explained before in 
>> https://trac.sagemath.org/ticket/32532#comment:11) in particular, Python 
>> users / developers generally have a working gcc (because many Python 
>> packages include extension modules that need to be compiled) but they have 
>> no need to have gfortran installed. The reason is that the few Python 
>> packages that use gfortran (such as scipy) have high-quality deployments of 
>> wheels to PyPI, which are widely used. Sage, however, does not yet have a 
>> working strategy that would enable us to use wheels from PyPI. As you know, 
>> I have been working on a PyPI-based deployment solution for Sage in 
>> https://trac.sagemath.org/ticket/29039; this builds everything from 
>> source. 
>>
>  
> Why does it build everything from source? E.g. numpy doesn't build Fortran 
> compilers.
> Why do we have to outdo other Python projects here?
>

I have explained in the paragraph that you quoted (and previously in 
#32532). Wheels. They have wheels. We don't.


 

-- 
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/bffb95ab-3407-487f-8e19-b8945f7e10c3n%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

2021-09-24 Thread Julien Puydt

Hi,

Le vendredi 24 septembre 2021 à 12:23 +0100, Dima Pasechnik a écrit :
> 
> On Fri, 24 Sep 2021, 02:12 Matthias Koeppe,
>  wrote:
> > I repeat my strong objections to the proposed change -- which does
> > not solve any problems and only creates new ones.
> > 
> > 1. When you refer to the "huge time sink", I think you are
> > referring to painful memories from some distant past. But we have
> > no current or recent problem with either the gcc or gfortran
> > package. I fixed the last problems with the gfortran package in
> > early 2020; there have not been any problems since.
> 
> I am referring to the time sink caused by the mere presence of
> gcc/gfortran spkg in Sage. Namely,
> 
> - support of outdated systems nobody uses (many tickets supporting
> past EOL systems recently)
> 
> - support of modern, but broken, systems (e.g. we just should not
> care that a particular Fedora version broke its gcc, it's not out
> problem, let us not try to bite more than we can chew)
> 
> - users who attempt to build gfortran as they don't have it installed
> - many fail, but I guess many more succeed,
> and rebuild it for each new release
> 
> - our Linux and  macOS binary builds are increasingly useless, and a
> timesink for everyone involved, potential users too.
> 
> > 

I made the point years ago that sagemath should be broken in two:

- sagemath-the-software ;

- sagemath-the-distribution.

This thread is an illustration of why that would simplify matters:

- sagemath developpers could focus on making the sofware work (both
increasing its scope/efficiency and solving real issues) ;

- the various distribution packagers would handle their specific
problems, and only complain upstream when there's an actual problem.

Cheers,

J.Puydt

-- 
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/701d2621e3dd450517b1cb4d572ab5f2df2587ac.camel%40gmail.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

2021-09-24 Thread Michael Orlitzky

On Thu, 2021-09-23 at 21:57 -0700, Matthias Koeppe wrote:
> On Thursday, September 23, 2021 at 9:09:45 PM UTC-7 Michael Orlitzky wrote:
> 
> > Item #2 happened, incidentally, only because we've copy & pasted so 
> > many packages into sage pretending to be a linux distribution. 
> > Eliminating that problem altogether begins right here, at the root of 
> > the dependency tree.
> > 
> 
> I don't think so. The packages that were not ready for gcc 11 support at 
> that time also were not ready upstream nor in other Linux distributions 
> (see https://trac.sagemath.org/ticket/31786). The issue is not so much that 
> we duplicate work by linux distributions but rather that Sage has 
> effectively adopted many poorly maintained or semi-abandoned math packages. 
> 

The list of blockers there implicates,

  * gcc
  * eclib
  * singular
  * ntl
  * gfortran
  * libgd
  * ecl
  * lcalc
  * fplll
  * giac

all of which are available on Fedora 34, and pretty much everywhere.
lcalc and fplll have the smallest number of entries in
build/pkgs/$pkg/distros, with seven.

While they may not all have been patched upstream (lcalc upstream was
gone), I'm fairly sure they were all patched in Gentoo or sage-on-
gentoo before they were patched in sage itself -- especially if I
change that wording to "patched in a release." I'd also bet that Conda,
Nix, and friends had it all sorted out before sage did. And all without
wasting sage developers' time (except for those of us who maintain the
distro packages, too, and who got to do everything twice).

We are only concerned about gcc-11 support in any of those packages
because we're supporting their SPKGs. But, for a while, the SPKGs have
had superior alternatives. Delete the SPKGs and you might run into some
other corner cases, but you certainly aren't going to hit gcc-11 build
failures if you aren't building.

-- 
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/e62326fe55846ad944bad815e68dd206e51047be.camel%40orlitzky.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Fri, 24 Sep 2021, 02:12 Matthias Koeppe, 
wrote:

> I repeat my strong objections to the proposed change -- which does not
> solve any problems and only creates new ones.
>
> 1. When you refer to the "huge time sink", I think you are referring to
> painful memories from some distant past. But we have no current or recent
> problem with either the gcc or gfortran package. I fixed the last problems
> with the gfortran package in early 2020; there have not been any problems
> since.
>

I am referring to the time sink caused by the mere presence of gcc/gfortran
spkg in Sage. Namely,

- support of outdated systems nobody uses (many tickets supporting past EOL
systems recently)

- support of modern, but broken, systems (e.g. we just should not care that
a particular Fedora version broke its gcc, it's not out problem, let us not
try to bite more than we can chew)

- users who attempt to build gfortran as they don't have it installed -
many fail, but I guess many more succeed,
and rebuild it for each new release

- our Linux and  macOS binary builds are increasingly useless, and a
timesink for everyone involved, potential users too.

> 2. We used the gcc spkg as recently as this Spring, for the 9.3 release,
> in order to support Fedora 34, which had just switched to gcc 11. No other
> solution was proposed or feasible at the time because several of our
> standard packages were not ready for this compiler, and a Sage release was
> already long overdue.
>

nobody pushes us into not skipping one particular version.
I cannot imagine e.g. cpython or scipy delaying a release in such a
situation.
Besides, building gcc from source is not too hard
with a modern C compiler available.

>
> 3. Your claims regarding availability of the tools on users' systems are
> not substantiated.
> a) clang/clang++, other than the version on macOS, is NOT supported by the
> Sage distribution. Only recently we added tests for these compilers (
> https://trac.sagemath.org/ticket/30835), which revealed build problems
> that are still unresolved (https://trac.sagemath.org/ticket/32207).
>

#32207 is resolved, and ready for review :-)

b) Any talk about FreeBSD is unsubstantiated, there is no systematic effort
> to support it or even test this environment.
>

There is a FreeBSD sagemath package: https://www.freshports.org/math/sage,
currently at 9.2.+patches.
It's of course much more fun to fool around with past-EOL linux systems,
just cause it's easier with GitHub Actions.

c) The use of any Fortran compiler other than homebrew's packaging of
> gfortran on macOS (and our gfortran spkg) is completely unexplored. Given
> the instability of homebrew -- as a rolling platform on which it is not
> possible to install a specific version, it would be reckless to tie
> ourselves to homebrew's gfortran as the only option. (An effort to support
> macports in https://trac.sagemath.org/ticket/31505 is stalled.)
>

In questionable cases homebrew does offer a choice of versions (e.g. with
python, or, indeed, gcc)
Besides, they are quick in fixing their problems, qucker than us.

As to other Fortran compilers' support - no, I did in the past look at
flang, and currently look at NAG's nagfor support.
Unfortunately flang isn't really considered production quality yet
(although perhaps it's getting there).
Good news about flang is that it's merged in the clang tree:
https://releases.llvm.org/11.0.0/tools/flang/docs/ReleaseNotes.html

d) Our build on top of conda has been broken for months.
>

So? I think it's good to burn the gcc/gfortan bridges, and instead fix the
conda build.

> 4. I want to caution about a single-user-system centric point of view.
> Non-root users on a multi-user unix system, or on a locked-down corporate
> laptop CANNOT install a system package. By removing SPKGs and demanding
> that the corresponding system package is available, we would make the Sage
> distribution harder to install for users, thus reduce the value that the
> Sage distribution provides for non-root users.
>

our experience dealing with requests for help from HPC clusters here is
that the ability to build gcc/gfortran
is mostly useless for them, as the systems tooling is old or just weird.

Locked down to the inability to install things corporate laptops running
Linux or macOS?
I've never seen one.

> 5. The gcc spkg is specifically used in building the binary distribution.
> Any discussion of removing the gcc/gfortran spkgs without including a
> discussion of deployment as the binary package is incomplete. Availability
> of the compilers at runtime is a key feature that we have advertised
> through facilities such as the magic %cython directive and others.
>

yet one cannot reliably (or not at all) install optional packages  in such
a scenario.
I consider the linux and macOS binary builds as mostly a timesink (see
above)

> 6. (As I explained before in
> https://trac.sagemath.org/ticket/32060#comment:36) There is a crucial
> distinction between the gcc and the

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thursday, September 23, 2021 at 9:09:45 PM UTC-7 Michael Orlitzky wrote:

> Item #2 happened, incidentally, only because we've copy & pasted so 
> many packages into sage pretending to be a linux distribution. 
> Eliminating that problem altogether begins right here, at the root of 
> the dependency tree.
>

I don't think so. The packages that were not ready for gcc 11 support at 
that time also were not ready upstream nor in other Linux distributions 
(see https://trac.sagemath.org/ticket/31786). The issue is not so much that 
we duplicate work by linux distributions but rather that Sage has 
effectively adopted many poorly maintained or semi-abandoned math packages. 

-- 
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/c1ea92a8-992a-4403-97cb-0598f0c235d6n%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thu, 2021-09-23 at 21:18 -0700, Matthias Koeppe wrote:
> On Thursday, September 23, 2021 at 7:48:35 PM UTC-7 Michael Orlitzky wrote:
> 
> > You could tar up an entire Gentoo system with 
> > sage installed and it would probably take up less space than our 
> > existing binaries. 
> > 
> 
> Gentoo is not relocatable, so that would not be useful.
> 

If the entire system is in the tarball, you can chroot right into the
extracted directory.




-- 
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/8cbc2957acbdd80caea1c0670340c8db87410a0b.camel%40orlitzky.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thursday, September 23, 2021 at 9:28:36 PM UTC-7 Michael Orlitzky wrote:

> On Thu, 2021-09-23 at 21:10 -0700, Matthias Koeppe wrote: 
> > Item 7 of my list, pip-installability, is definitely not addressed by 
> any 
> > of these. From a Python dev viewpoint, a project that is not 
> > pip-installable does not exist. 
> > 
>
> The only concrete complaint I saw in #7 was "Python users would again 
> need root access to install a system package to get [gfortran]." 
>
> I admit to having very little understanding of your larger plan for 
> pip-installability and modularizing sagelib. I've read all of the 
> comments about it, but I still don't see what the gfortran SPKG has to 
> do with it.


The ticket https://trac.sagemath.org/ticket/29039 explains it.
gfortran is just one of the packages that is part of the 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+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/aab654b2-e4e2-4959-888a-e22f5e54b726n%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thu, 2021-09-23 at 21:10 -0700, Matthias Koeppe wrote:
> On Thursday, September 23, 2021 at 7:48:35 PM UTC-7 Michael Orlitzky wrote:
> 
> > 2, 4, 6, and 7 are addressed by Conda, Nix, Guix, Homebrew, or even 
> > Gentoo Prefix. 
> > 
> 
> Item 7 of my list, pip-installability, is definitely not addressed by any 
> of these. From a Python dev viewpoint, a project that is not 
> pip-installable does not exist. 
> 

The only concrete complaint I saw in #7 was "Python users would again
need root access to install a system package to get [gfortran]."

I admit to having very little understanding of your larger plan for
pip-installability and modularizing sagelib. I've read all of the
comments about it, but I still don't see what the gfortran SPKG has to
do with it.


-- 
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/8ce33adece22563d3250bca86ae28dcfebaeabc0.camel%40orlitzky.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thursday, September 23, 2021 at 7:48:35 PM UTC-7 Michael Orlitzky wrote:

> You could tar up an entire Gentoo system with 
> sage installed and it would probably take up less space than our 
> existing binaries. 
>

Gentoo is not relocatable, so that would not be useful. gentoo-prefix is 
hardly production-ready last time I checked 
(https://trac.sagemath.org/ticket/29105), and there is no evidence anyone 
has ever built Sage on it.

-- 
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/587265f3-6b83-41a5-b60f-c0bdc5c1ce10n%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thu, 2021-09-23 at 20:51 -0700, Matthias Koeppe wrote:
> On Thursday, September 23, 2021 at 7:48:35 PM UTC-7 Michael Orlitzky wrote:
> 
> > Whoever is building the binary images could install a compiler into 
> > SAGE_LOCAL with "dpkg-deb -x" or whatever before they start. There's no 
> > need for it to come from an SPKG.
> 
> 
> I don't think Debian packages are relocatable. "dpkg-deb -x" is 
> specifically NOT for installing, but for inspecting.
> So there goes another suggestion to replace something that works by 
> something conjectural.
> 
> 

"or whatever" was meant to suggest that if this exact command won't
work, then something similar to it will. #5 is a legitimate problem,
but it's not a hard one.


-- 
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/20ac169d9c976f21abb3932e5877476aa23fe05c.camel%40orlitzky.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thursday, September 23, 2021 at 7:48:35 PM UTC-7 Michael Orlitzky wrote:

> We ship a 
> 3GB image of an entire operating system. This could be done much easier 
> by... shipping an image of an entire operating system: QEMU, 
> VirtualBox, Docker, etc. 
>

We don't even seem to have enough developer interest in this to keep the 
existing Docker build of Sage in good shape.

-- 
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/cf94a176-beae-4e11-949e-b109753f8f5fn%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thursday, September 23, 2021 at 7:48:35 PM UTC-7 Michael Orlitzky wrote:

> 2, 4, 6, and 7 are addressed by Conda, Nix, Guix, Homebrew, or even 
> Gentoo Prefix. 
>

Item 7 of my list, pip-installability, is definitely not addressed by any 
of these. From a Python dev viewpoint, a project that is not 
pip-installable does not exist. 

-- 
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/f422d51b-1ba0-4ada-81c5-7ed7437f6f85n%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thu, 2021-09-23 at 20:32 -0700, John H Palmieri wrote:
> Item #2 actually happened. Can you cite posts from people successfully 
> using the tools you mention to get around it? 

People who solve the unnecessary problem in the obvious and reliable
way don't post about it. If you really doubt that these tools can be
used to work around it, you can download sage-9.3 on fedora-34 and try
it. But I don't think anyone is claiming that they don't work.

Item #2 happened, incidentally, only because we've copy & pasted so
many packages into sage pretending to be a linux distribution.
Eliminating that problem altogether begins right here, at the root of
the dependency 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+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/5c46d1766dd2c49b5b17df4a4be60847f0f20c52.camel%40orlitzky.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thursday, September 23, 2021 at 7:48:35 PM UTC-7 Michael Orlitzky wrote:

> Whoever is building the binary images could install a compiler into 
> SAGE_LOCAL with "dpkg-deb -x" or whatever before they start. There's no 
> need for it to come from an SPKG.

I don't think Debian packages are relocatable. "dpkg-deb -x" is 
specifically NOT for installing, but for inspecting.
So there goes another suggestion to replace something that works by 
something conjectural.

-- 
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/50c6dcaa-32c6-49b4-a0f5-99488b5fde0fn%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

2021-09-23 Thread John H Palmieri

Item #2 actually happened. Can you cite posts from people successfully 
using the tools you mention to get around it? 


On Thursday, September 23, 2021 at 7:48:35 PM UTC-7 Michael Orlitzky wrote:

> On Thu, 2021-09-23 at 18:12 -0700, Matthias Koeppe wrote:
> > I repeat my strong objections to the proposed change -- which does not 
> > solve any problems and only creates new ones.
> > 
>
> 2, 4, 6, and 7 are addressed by Conda, Nix, Guix, Homebrew, or even
> Gentoo Prefix. These projects *want* to build entire distributions that
> run out of your home directory.
>
>
>
> > 5. The gcc spkg is specifically used in building the binary 
> distribution. 
> > Any discussion of removing the gcc/gfortran spkgs without including a 
> > discussion of deployment as the binary package is incomplete. 
> Availability 
> > of the compilers at runtime is a key feature that we have advertised 
> > through facilities such as the magic %cython directive and others.
> > 
>
> Whoever is building the binary images could install a compiler into
> SAGE_LOCAL with "dpkg-deb -x" or whatever before they start. There's no
> need for it to come from an SPKG.
>
> The fact that we even have to consider these issues supports the claim
> that these packages are a maintenance burden.
>
> And, veering off-topic:
>
> Our binaries aren't really binaries in the traditional sense. We ship a
> 3GB image of an entire operating system. This could be done much easier
> by... shipping an image of an entire operating system: QEMU,
> VirtualBox, Docker, etc. You could tar up an entire Gentoo system with
> sage installed and it would probably take up less space than our
> existing binaries. Similarly, I'm sure we could tar up a Conda image
> where sage was installed and tell people to download that. Provide a
> launcher script and it would work out-of-the-box.
>
>
>

-- 
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/fddc13dc-a843-4bc2-a06e-390a54cbe0fan%40googlegroups.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

2021-09-23 Thread William Stein

On Thu, Sep 23, 2021 at 7:48 PM Michael Orlitzky  wrote:
> On Thu, 2021-09-23 at 18:12 -0700, Matthias Koeppe wrote:
> And, veering off-topic:
>
> Our binaries aren't really binaries in the traditional sense. We ship a
> 3GB image of an entire operating system. This could be done much easier
> by... shipping an image of an entire operating system: QEMU,
> VirtualBox, Docker, etc. You could tar up an entire Gentoo system with
> sage installed and it would probably take up less space than our
> existing binaries. Similarly, I'm sure we could tar up a Conda image
> where sage was installed and tell people to download that. Provide a
> launcher script and it would work out-of-the-box.

(Staying off topic.)

I installed Isuru Fernando's Conda image of SageMath on my Macbook Pro
M1 "Apple Silicon" yesterday,
and it worked quickly with no trouble that I could see.   Here's how
big the install is (decompressed):

wstein@Williams-MBP ~ % du -sch sagemath-forge
5.5G sagemath-forge
5.5G total

Compilers are included as part of this, so Cython does just work out
of the box as far as I can tell:

sage: cython('def f(n): return n+1')sage: f(10)
11
sage: !which clang
/Users/wstein/sagemath-forge/bin/clang
sage: !which gfortran
/Users/wstein/sagemath-forge/bin/gfortran

In any case, the momentum behind conda-forge is really impressive at
this point... though this particular
page on their website

https://conda-forge.org/feedstock-outputs/

listing all packages is really horrible, since it slowly loads a
massive list in random order, etc.   It's funny
that it feels so broken as a result of them being so successful and
getting contributions!  (It looks like there
might be around 14K packages listed there?)

Isuru was explaining to me earlier today that the bug that have caused
some trouble with
the sage-9.3 and sage-9.4 binaries are due to OpenBLAS build options.
The conda-forge maintainers also
hit these same problems with their openblas package (which sage in
conda just depends on), and
fixed that one single package a while ago.   (DISCLAIMER: I'm
repeating this second hand so may
have some details wrong. )  In any case, conda seems to be doing a
surprisingly good job at solving
a really, really hard problem.

 -- William

>
>
> --
> 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/381f2f471070a1192a1b7ba79799bfb3eee1acbd.camel%40orlitzky.com.

-- 
William (http://wstein.org)

-- 
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/CACLE5GAvvjbh8aaRzQfSCccq38r%2B3VSWF4S36E-axH6p8_Ji2w%40mail.gmail.com.

Re: [sage-devel] Re: #32532 - removing gcc and gfortran spkgs

On Thu, 2021-09-23 at 18:12 -0700, Matthias Koeppe wrote:
> I repeat my strong objections to the proposed change -- which does not 
> solve any problems and only creates new ones.
> 

2, 4, 6, and 7 are addressed by Conda, Nix, Guix, Homebrew, or even
Gentoo Prefix. These projects *want* to build entire distributions that
run out of your home directory.

> 5. The gcc spkg is specifically used in building the binary distribution. 
> Any discussion of removing the gcc/gfortran spkgs without including a 
> discussion of deployment as the binary package is incomplete. Availability 
> of the compilers at runtime is a key feature that we have advertised 
> through facilities such as the magic %cython directive and others.
> 

Whoever is building the binary images could install a compiler into
SAGE_LOCAL with "dpkg-deb -x" or whatever before they start. There's no
need for it to come from an SPKG.

The fact that we even have to consider these issues supports the claim
that these packages are a maintenance burden.

And, veering off-topic:

Our binaries aren't really binaries in the traditional sense. We ship a
3GB image of an entire operating system. This could be done much easier
by... shipping an image of an entire operating system: QEMU,
VirtualBox, Docker, etc. You could tar up an entire Gentoo system with
sage installed and it would probably take up less space than our
existing binaries. Similarly, I'm sure we could tar up a Conda image
where sage was installed and tell people to download that. Provide a
launcher script and it would work out-of-the-box.

-- 
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/381f2f471070a1192a1b7ba79799bfb3eee1acbd.camel%40orlitzky.com.

[sage-devel] Re: #32532 - removing gcc and gfortran spkgs