cython_aliases: Use ecl-config to determine compiler/linker flags for ecl

[mkoeppe]

We use ecl-config to determine compiler/linker flags for ecl. This is a step towards using a system-installed ecl and is needed/helpful for installing sage in virtual environments (#30371).

$ ecl-config --help
Usage: /usr/local/bin/ecl-config [OPTIONS] [LIBS]
Options: 
	[--cflags]
	[--libs|--ldflags]
Libs:
	cmp

Sample output:

> ./local/bin/ecl-config --cflags --libs
-Dlinux   -I/mnt/d/Programming/sage/local/include
-Wl,--rpath,/mnt/d/Programming/sage/local/lib -L/mnt/d/Programming/sage/local/lib -lecl -Wl,-rpath-link,/mnt/d/Programming/sage/local/lib -L/mnt/d/Programming/sage/local/lib -Wl,-rpath,/mnt/d/Programming/sage/local/lib      -ldl -lm

[mkoeppe]

Ideally, of course, ecl would install a pkg-config module...

[mkoeppe]

Hoping we can make progress on this ticket this week - ​https://wiki.sagemath.org/days111

[gh-tobiasdiez]

The following is working for me (in the context of #30371).

---

New commits:

​3049e53

Use ecl-config to determine compiler/linker flags for ecl

[mkoeppe]

Following the conventions of autoconf, this variable should be called ECL_CONFIG, not ECL_CONFIG_PATH

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​1041128

Rename ecl config variable

[gh-tobiasdiez]

Replying to mkoeppe:

Following the conventions of autoconf, this variable should be called ECL_CONFIG, not ECL_CONFIG_PATH

I'm not sure if one should follow the autoconf conventions here since the variable is (right now) only used in python scripts, but I've renamed it anyway.

[mkoeppe]

capture_output is not available on Python 3.6 - see #30758

[mkoeppe]

I think it's better to make the simple change of the code as in ​https://git.sagemath.org/sage.git/commit?id=a569b88f02ab4dfdd8eec6e8ad0472f1f537d2c8 instead of deferring this ticket until after we drop support for python 3.6

[dimpase]

I have yet to see a system that ships patched ECL, allowing one to build Maxima in Sage, etc.

[gh-tobiasdiez]

I use this ticket successfully in #30371 to build sage in a new python virtual environment using sage's ecl precisely because the system ecl is not working.

[mkoeppe]

Replying to dimpase:

I have yet to see a system that ships patched ECL, allowing one to build Maxima in Sage, etc.

Surely distributions that ship sage must have a suitable package: conda, arch, ...?

[arojas]

@@ -203,6 +204,7 @@ var('MAXIMA_FAS')
 var('SAGE_NAUTY_BINS_PREFIX',        '')
 var('ARB_LIBRARY',                   'arb')
 var('CBLAS_PC_MODULES',              'cblas:openblas:blas')
+var('ECL_CONFIG')
 
 # misc
 var('SAGE_BANNER', '')

Please add a fallback here, otherwise this will break distro packaging without sage_conf

[arojas]

Given that sage's maxima is patched to install a fas module in ecl's tree, how is this supposed to work with a system ecl?

[mkoeppe]

I think we can change our maxima installation to use a different installation directoery and then pass the full path using the existing variable sage.env.MAXIMA_FAS

[mkoeppe]

Replying to arojas:

@@ -203,6 +204,7 @@ var('MAXIMA_FAS')
 var('SAGE_NAUTY_BINS_PREFIX',        '')
 var('ARB_LIBRARY',                   'arb')
 var('CBLAS_PC_MODULES',              'cblas:openblas:blas')
+var('ECL_CONFIG')
 
 # misc
 var('SAGE_BANNER', '')

Please add a fallback here, otherwise this will break distro packaging without sage_conf

I agree, this should fall back to just 'ecl-config'

[gh-tobiasdiez]

Good catch, I've now added a fallback!

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​fd05a97

Add fallback

[dimpase]

#30551 is for 9.4, why is this one still for 9.3 ? (I'd rather switch #30551 to 9.3, though)

[gh-tobiasdiez]

I've put #30551 now to 9.3 since I agree it would be convenient to drop python 3.6 support soon. Matthias, please feel free to put both tickets onto 9.4 if you strongly feel that 3.6 still needs to be supported.

[mkoeppe]

Let's please not use a ticket like this where we know exactly how to write python 3.6 conforming code to force a decision that needs proper discussion in #30551.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​6370c34

src/sage/env.py: Do not use capture_output, which requires python 3.7

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​31f1bbb

Also use universal_newlines instead of text, for python 3.6 compatibility

[dimpase]

Replying to git:

Branch pushed to git repo; I updated commit sha1. New commits:

​31f1bbb

Also use universal_newlines instead of text, for python 3.6 compatibility

Why do you push Python-3.6-needed changes to a ticket that depends on #30551 ? It just makes no sense to me.

[mkoeppe]

Thanks. I forgot to remove the dependency.

[mkoeppe]

Replying to arojas:

Given that sage's maxima is patched to install a fas module in ecl's tree, how is this supposed to work with a system ecl?

Where do distributions install compiled common lisp libraries?

[fbissey]

Replying to mkoeppe:

Replying to arojas:

Given that sage's maxima is patched to install a fas module in ecl's tree, how is this supposed to work with a system ecl?

Where do distributions install compiled common lisp libraries?

Gentoo does it in /usr/lib64/ecl-20.4.24 - as can be seen it is versioned. This is the place where ecl put a lot of its own libraries. It is determined by running the following ecl lisp formula, which I believe is the same in vanilla sage.

ecl -eval "(princ (SI:GET-LIBRARY-PATHNAME))" -eval "(quit)"

It could be interesting to know other, or how to set other, paths to install .fas files for ecl.

[gh-spaghettisalat]

Replying to fbissey:

Replying to mkoeppe:

Replying to arojas:

Given that sage's maxima is patched to install a fas module in ecl's tree, how is this supposed to work with a system ecl?

Where do distributions install compiled common lisp libraries?

Gentoo does it in /usr/lib64/ecl-20.4.24 - as can be seen it is versioned. This is the place where ecl put a lot of its own libraries. It is determined by running the following ecl lisp formula, which I believe is the same in vanilla sage.

ecl -eval "(princ (SI:GET-LIBRARY-PATHNAME))" -eval "(quit)"

It could be interesting to know other, or how to set other, paths to install .fas files for ecl.

There is no universally agreed place to put .fas files, after all you can just use (load "/wherever/whatever.fas") to load them from any place you like. But /usr/lib64/ecl-20.4.24 makes the most sense for libraries useable for multiple projects and also allows you to call (require :whatever) to use the fas in ECL. On the other hand, if the .fas file is only useable for a single project, I would rather put it in /usr/lib64/some-other-project/ and explicitely load it from there.

[fbissey]

This is extremely interesting, how would you load a .fas from an arbitrary location?

[gh-spaghettisalat]

Replying to fbissey:

This is extremely interesting, how would you load a .fas from an arbitrary location?

As I said, you just call (load "/path/to/lib.fas"). This is a standard common lisp function for loading compiled code or source code into the running image, much like dlopen loads a shared library in C.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​2ac4cc8	Merge branch 'develop' of git://github.com/sagemath/sage into public/build/ecl-config
​11ad40a	Merge branch 'public/build/ecl-config' of git://trac.sagemath.org/sage into public/build/ecl-config

[mkoeppe]

Let's continue the discussion on Maxima FASL installation in #29617 (spkg-configure.m4 and distros/ information for ecl)

The present ticket is about the parameters for compiling the sage.libs.ecl extension. Needs review

[mkoeppe]

Let's please get this in

[dimpase]

lgtm

[gh-tobiasdiez]

Thanks!

[mkoeppe]

Follow up in #33803