Sort out linking with libstdc++

[dimpase]

From the post at

• https://groups.google.com/g/sage-devel/c/YUGe_Ey823Y/m/zkjpE-YYBgAJ

it appears linking with libstdc++ might be not needed any more, and so this may be simplified.

Note that clang project has libc++, which serves the same role, but is a different implementation.

(Currently stdc++ is mentioned in src/setup.py
and in src/sage/misc/cython.py.)

Depends on #29702

CC: @antonio-rojas @mkoeppe @kiwifb @isuruf

Component: build

Author: Dima Pasechnik

Branch/Commit: afeec95

Reviewer: Matthias Koeppe

Issue created by migration from https://trac.sagemath.org/ticket/29856

[dimpase]

comment:1

trying to see what happens with the simple

index af5f99d03e..0600603e62 100644
--- a/src/sage/misc/cython.py
+++ b/src/sage/misc/cython.py
@@ -41,7 +41,7 @@ cblas_library_dirs = list(cblas_pc['library_dirs'])
 cblas_include_dirs = list(cblas_pc['include_dirs'])
 
 standard_libs = [
-    'mpfr', 'gmp', 'gmpxx', 'stdc++', 'pari', 'm',
+    'mpfr', 'gmp', 'gmpxx', 'pari', 'm',
     'ec', 'gsl',
 ] + cblas_libs + [
     'ntl']
--- a/src/setup.py
+++ b/src/setup.py
@@ -354,8 +354,6 @@ class sage_build_cython(Command):
 
         - Add dependencies on header files for certain libraries
 
-        - Ensure that C++ extensions link with -lstdc++
-
         - Sort the libraries according to the library order
 
         - Add some default compile/link args and directories
@@ -369,10 +367,7 @@ class sage_build_cython(Command):
         lang = kwds.get('language', 'c')
         cplusplus = (lang == "c++")
 
-        # Libraries: add stdc++ if needed and sort them
         libs = kwds.get('libraries', [])
-        if cplusplus:
-            libs = libs + ['stdc++']
         kwds['libraries'] = sorted(set(libs),
                 key=lambda lib: library_order.get(lib, 0))

[mkoeppe]

comment:3

This ticket description needs some work; what problem is this ticket addressing?

Also, setup.py does not use sage.misc.cython at all if I'm not mistaken.

And please note #29702 moves the code from src/setup.py to sage_setup - best to do changes on top of that

[dimpase]

Author: Dima Pasechnik

[dimpase]

Commit: 07010d4

[dimpase]

Dependencies: #29702

[dimpase]

Branch: u/dimpase/build/get_rid_of_stdcxx

[dimpase]

comment:6

seems to work.

---

Last 10 new commits:

5790687	build/bin/write-dockerfile.sh: Do not ADD removed file src/Makefile.in
851ab76	Make sagelib a script package
f39a017	Move src/fpickle_setup.py to srs/sage_setup/
363f792	src/setup.py: Move class sage_install to new module sage_setup.command.sage_install
a7cde9c	src/module_list.py: Move library_order to new module sage_setup.library_order
7c4bbf6	src/setup.py: Move classes sage_build_* to new modules sage_setup.command.sage_*
ddc49aa	src/setup.py: Move excepthook to new module sage_setup.excepthook
9cff0c6	src/sage_setup/command/__init__.py: New
56705f8	src/sage_setup/command/sage_build_cython.py: Fix adding setup.py as dependency
07010d4	remove stdc++ from everywhere

[mkoeppe]

comment:7

tested on GH Actions?

[mkoeppe]

comment:8

Please rebase on unrebased-then-merged #29702

[dimpase]

Changed branch from u/dimpase/build/get_rid_of_stdcxx to u/dimpase/dont_link_to_stdcxx

[dimpase]

Changed commit from 07010d4 to afeec95

[dimpase]

Last 10 new commits:

e810ad1	Trac #29345: don't use sage's config.status for the lrcalc build.
93c9921	Trac #29345: replace the function that populates the CVXOPT_* variables.
0e66a0a	Trac #29345: add Dima's SPKG patches for ksh compatibility.
df3f05e	build/make/Makefile.in [SCRIPT_PACKAGE_templ]: cd into the SPKG directory; adjust spkg-install scripts
5372065	Merge branch 't/29793/script_packages_should_cd_into_the_spkg_directory' into t/29411/make_sagelib_a_script_package
c166b97	Merge branch 't/29411/make_sagelib_a_script_package' into t/29702/move_all_code_from_src_setup_py__src_fpickle_setup_py_to_sage_setup
cc30471	build/bin/write-dockerfile.sh: Do not ADD removed file src/Makefile.in
8a41326	Merge branch 't/29411/make_sagelib_a_script_package' into t/29702/move_all_code_from_src_setup_py__src_fpickle_setup_py_to_sage_setup
a56dc35	Merge tag '9.2.beta1' into t/29702/public/move_all_code_from_src_setup_py__src_fpickle_setup_py_to_sage_setup
afeec95	explictly linking with libstdc++ not needed

[dimpase]

comment:10

testing on https://github.com/dimpase/sage/pull/10 (also, this will test #29702)

[mkoeppe]

comment:11

Great

[dimpase]

comment:12

Replying to @dimpase:

testing on https://github.com/dimpase/sage/pull/10 (also, this will test #29702)

I see a lot of assembler errors (GNU 'as' version 2.24, ~6 years old)
in building ecm, and elsewhere, on minimal ubuntu-trusty

2020-06-16T00:41:24.9539407Z   [ecm-7.0.4.p1]   /tmp/ccEhz4KE.s:342: Error: operand size mismatch for `vmovdqu64'

see https://github.com/dimpase/sage/runs/774196885

Not a surprise, as we build a newish gcc, not really compatible with an old as, running on hardware not 100% supported by this as. Should we just drop this target all together?

[dimpase]

comment:13

On ubunty-bionic standard I see

2020-06-16T01:19:21.1467385Z   [dochtml]   ImportError: /usr/lib/x86_64-linux-gnu/libstdc++.so.6: version `GLIBCXX_3.4.26' not found (required by /sage/local/lib/python3.7/site-packages/sage/matrix/matrix_integer_sparse.cpython-37m-x86_64-linux-gnu.so)

which looks like a broken toolchain (`GLIBCXX_3.4.26' not found) - any idea why?

see https://github.com/dimpase/sage/runs/774196977

The rest looks OK to me.

[mkoeppe]

comment:14

The above two, are these from -gcc_spkg builds? Please add links to the runs

[mkoeppe]

comment:15

(See #29675 for known failures of -gcc_spkg builds with "GLIBCXX not found" errors.)

[dimpase]

comment:16

Replying to @mkoeppe:

The above two, are these from -gcc_spkg builds? Please add links to the runs

done

[mkoeppe]

Reviewer: Matthias Koeppe

[vbraun]

Changed branch from u/dimpase/dont_link_to_stdcxx to afeec95