Meta-ticket: Make sage (the distribution) behave like a standard autotools package, to the extent possible

[mkoeppe]

The goal of this ticket is to be able to explain to experienced Unix users what sage-the-distribution is, as follows:

If you download sage-the-distribution and do

./configure --prefix=/SOMEWHERE && make 

then that's the same as doing a sequence of about 100 times:

wget SOMEPACKAGE.tar.gz && tar xf SOMEPACKAGE.tar.gz && (cd SOMEPACKAGE \
&& ./configure --prefix=/SOMEWHERE && make && make install)

(and about a 50 times pip install SOMEPACKAGE). sage-the-distribution has figured out the right order of installing these packages, tricky configure options so that everything works, and on top has fixes for various outdated/handwritten/missing build systems of various packages.

Thus it is beyond the scope of this ticket:

• to separate make and make install. Our make install is a no-op. ANY discussion of this needs to go to ticket #21495, not here.

We will implement this goal without sacrificing any of the traditional convenience features that Sage-the-distribution has provided for the casual user (such as sage -i for installing packages; and that ./configure is an optional step of the installation process).

We have a separate task ticket for the following:

• #21507: Make sagelib (sage-the-Python-library) a normal Python package, installable by setup.py, pip, etc. and eventually even via PyPI. We will NOT turn sagelib to an automake package (which was proposed in #14807).

(There will be some interaction with some of the steps of that ticket.)

Included on this ticket are the following steps.

Implement standard features expected of an autotools build system.

• Choosing the installation hierarchy (configure --prefix=SAGE_LOCAL). Right now it is the subdirectory local of SAGE_ROOT.
  • #21501: Allow SAGE_LOCAL to be customized
  • #21534: Allow SAGE_LOCAL to be customized - follow-up
  • #21479: ./configure --prefix=SAGE_LOCAL
• By allowing the user to choose the installation hierarchy, there are new requirements. What is installed there should run without requiring environment variables to be set. It should not refer to the environment variables SAGE_LOCAL, or SAGE_ROOT (or SAGE_SRC_ROOT). An exception could perhaps be made for the latter for "debugging" or "source inspection" facilities:
  • #21525: package autotools: Don't depend on $SAGE_LOCAL
  • #21509: Install cython_debug somewhere in SAGE_LOCAL
• #21539: make V=0 should silence the build
• #21775: make distclean: Don't delete the install hierarchy (SAGE_LOCAL)
• #21591: Replace use of SAGE_ROOT by more specific environment variables
• #21469: Enable VPATH builds (SAGE_SRC_ROOT/configure --srcdir=SAGE_SRC_ROOT)
• #22646: ./configure CC=/path/to/gcc ...
• #28095: Add --enable-OPTIONALSPKG options to configure (see also earlier #21538)
• #29113: Reimplement sage -i SPKG for optional/experimental packages as configure --enable-SPKG && make build (follow-up: #29626)
• #21589: Make make V=0 the default by ./configure --enable-silent-rules
• Various other configure options, to replace use of environment variables that influence the build.

Clean up parts of the build system to make it more standard. This is to make it straightforward for developers familiar with the autotools system to contribute to sage.

• #21532: Create SAGE_LOCAL directory hierarchy during make, not configure
• #21524: configure.ac: write build/make/Makefile within an AC_CONFIG_COMMANDS, not during main configure

Make the separation between sage-the-distribution and sagelib (sage-the-Python-library) clearer. This will be beneficial for distributions such as Debian etc.

• #21559: Move sage-the-distribution scripts from src/bin to build/bin
• #21565: Add src/README.txt and build/README.txt

Following are workarounds to enable root-owned installation hierarchies (prefix).

• #21537: If $SAGE_SUDO is set, use it whenever we do make install of a package. Subtickets:
  • #21726: Support splitting spkg install into spkg-build and spkg-install (for SAGE_SUDO)
• #21536: documentation: Recommend GNU stow to people who want to install to /usr/local

See also:

• #29133 META-META-TICKET: Build, packaging, testing improvements

[mmezzarobba]

Do you intend ./configure to detect already installed dependencies so that make skips installing them? (Eventually? as part of this task?) For example, is the idea that ./configure would finish by saying something like “the following dependencies are missing: ... please install them, or type make and I'll install my own copy”? (This wouldn't be quite standard, but close enough, while staying mostly compatible with the installation procedure of sage-the-distribution.)

Or would sage-the-distribution still always have its private version of everything, so that users would have to choose between installing it and installing sagelib directly (e.g., via pip)?

[embray]

I wonder if this work might not also benefit from adopting portions of ​astropy-helpers into Sage's build tools. I don't think it would necessarily be appropriate to use directly--although there's nothing very "astropy" specific in astropy-helpers, a lot of it does assume one is using the astropy project template, or something close to it. But large swaths of it are very generic, and I put a lot of work into developing tools for build sanity of large Python projects.

Some key features that I think might benefit sage include:

• per-subpackage setup_package.py modules, which can contain any number of a ​(mostly) well-documented hooks for building compiled modules in that sub-package. This provides a clear and explicit alternative to a monolithic module_list.py (see the discussion ​here), and would also make it easier to break off some of Sage's subpackages into standalone packages should the time ever come for that.
• one of the less-well-documented features is also that setup_package.py can contain hooks for setup.py commands. For example, if a module contains some generated code (a la sage_setup.autogen), it can include a pre_build_ext hook to make sure the generated code is up to date before running the ./setup.py build_ext command (either explicitly, or implicitly as a sub-command). This makes it relatively easy for individual submodules to put hooks into the build process without having to write complicated command subclasses cluttered with subpackage-specific crud.
• smarter handling of Cython modules--for example, when creating a source tarball it will make sure all Cython modules have been Cythonized, and it will include the resulting C/C++ sources in the source tarball. This ensures that the user does not need Cython to be installed in order to build from source (which is good, because even if they do have Cython it will often be the wrong version--and in Sage's case missing important patches as well).

Those are the main ones for now but I could probably think of more.

[mkoeppe]

Replying to mmezzarobba:

Do you intend ./configure to detect already installed dependencies so that make skips installing them? (Eventually? as part of this task?)

That is beyond the scope of this task ticket. But I would be interested in a follow-up ticket to this effect.

[embray]

Will this also fix how make actually runs configure for you, and that configure gets invoked even if you run make clean targets?

[mkoeppe]

Replying to embray:

Will this also fix how make actually runs configure for you,

No, probably not. For this ticket, I want to keep configure an optional step of the installation process. (I've updated the description.)

and that configure gets invoked even if you run make clean targets?

Not sure about this one. To keep this ticket manageable, I want to change the top-level Makefile as little as possible. But please do open a ticket that describes make clean issues that ought to be addressed.

[embray]

Replying to mkoeppe:

Replying to embray:

Will this also fix how make actually runs configure for you,

No, probably not. For this ticket, I want to keep configure an optional step of the installation process. (I've updated the description.)

I guess it would be fine to make it an optional step. But I don't think it should be a mandatory step for all make targets, and currently it is, or at least seems to be. For example if I run make maintainer-clean twice in a row, the second time will run configure only to delete its output.

This is especially annoying on Cygwin where configure is extremely slow.

[jdemeyer]

Replying to embray:

Will this also fix how make actually runs configure for you, and that configure gets invoked even if you run make clean targets?

In standard autotools packages, the Makefile gets created by configure, so you wouldn't be able to run make clean either without configure.

[embray]

True--in that case it should (hopefully) be a moot point. This is currently a huge annoyance for me.

[embray]

Do we have a ticket somewhere for adding more configure-time checks for system packages? If there is I can't find it.

It's a little outside the scope of this ticket but not entirely. There are quite a few system packages installed by sage that could be skipped if we added the appropriate scripts to check for them.

[jdemeyer]

Replying to embray:

Do we have a ticket somewhere for adding more configure-time checks for system packages? If there is I can't find it.

No, we don't. Feel free to open a ticket. I would prefer one ticket for every package, not a single ticket trying to do too much.

[embray]

I agree--I probably won't even make a ticket for every package right away but I will make a master ticket for coordinating the task, and then open individual tickets for packages that I think are most worth tackling.

[mkoeppe]

Would someone who has followed recent build system changes be interested in updating the description of this meta-ticket?

[embray]

Replying to mkoeppe:

Would someone who has followed recent build system changes be interested in updating the description of this meta-ticket?

Which build system changes specifically? I have a few open tickets that are making some not insignificant changes, but they don't terribly change things for this ticket. If anything they will help make it easier to implement the goals of this ticket. There is one small conflict I can see: The work I'm doing in #22509 and in #23160 will mostly supersede the work you already did in #21537. This is because the sdh_make_install helper function I added will also handle $SAGE_SUDO.

Right now I'm working toward the goals I outlined ​here for improving support for building Sage against dependencies from the system. However, I haven't even made tickets for every aspect of that plan yet. All the work I'm currently doing is toward #22509 and #22510 which I see as necessary for saner package management in Sage (they will also be very helpful for work I need to do of making it easier to install optional packages with the Windows installer, but that's otherwise an orthogonal issue).