Accept gcc 11 (for fedora 34)

[Matthias Köppe]

After having resolved the following blockers:

• #29703 Upgrade gcc to 10.3 (latest 10.x), fix --without-system-gcc builds
• #31624 fedora-34-standard: eclib, singular build failures with system ntl
• #31838 gfortran: Accept system gfortran 11
• #31879/#27901 libgd
• #31694 pynac fails to compile with gcc-11 (headers porting) ​https://github.com/pynac/pynac/pull/375
• #32093 ecl: Add SAGE_SPKG_DEPCHECK for gcc
• #31837 lcalc headers prevent sage compilation ​https://github.com/cschwan/sage-on-gentoo/issues/634 and ​https://github.com/cschwan/sage-on-gentoo/blob/master/sci-mathematics/lcalc/files/lcalc-1.23-gcc11.patch
• #31832 fplll fails to compile with gcc-11 (headers porting) ​https://github.com/fplll/fplll/pull/463
• gcc 10 build fails in some circumstances on Fedora 34 (​https://groups.google.com/g/sage-devel/c/LcQ3ufUOJpI/m/lOsDMZ0qBQAJ, ​https://gcc.gnu.org/pipermail/gcc-help/2021-June/140439.html)

... we accept gcc 11 again. This will allow building Sage with the default compilers on Fedora 34 and other platforms.

Not so urgent:

• giac (any versions) doesn't compile with std=c++17 ​https://bugs.gentoo.org/788283
• Optional package latte_int doesn't compile with std=c++17 (exception handling) ​https://github.com/latte-int/latte/issues/24

[François Bissey]

I didn't have any problems in Gentoo with eclib and ntl.

[François Bissey]

Note that issue of standard do not arise in vanilla sage since we enforce std=c++11. However, issues with headers are not prevented by sticking to that standard, which means the following have to be fixed in any case:

• fplll
• pynac
• lcalc

Others can be delayed.

[Matthias Köppe]

New commits:

​2e8fe53	build/pkgs/{gcc,gfortran}/spkg-configure.m4: Accept gcc 11.x
​956ddcc	build/pkgs/fplll: Update to 5.4.1
​b2ea24e	build/pkgs/fpylll: Update to 0.5.6
​5486f81	Merge #31832

[Enrique Artal Bartolo]

The compilation failed with lcalc and pynac. The other packages in the introduction have been built, except for fplll which did not appear.

[Volker Braun]

Workaround for libgd is at #31879

[Dima Pasechnik]

rebase please

[Matthias Köppe]

The branch is no longer relevant, all tickets have been merged

[Dima Pasechnik]

By the way, a viable route to work on Fedora 34 might be to use clang rather than gcc.

dnf install clang
CC=clang CXX=clang++ ./configure...

Do we test Linux/clang on GH Actions? We should...

[Matthias Köppe]

Replying to dimpase:

Do we test Linux/clang on GH Actions? We should...

I agree. #30835 can be broadened to include this...

[Dima Pasechnik]

we see Maxima builds failing in Fedora 34 on the latests CI runs, see e.g. ​https://github.com/sagemath/sagetrac-mirror/runs/2918514130

[Matthias Köppe]

earlier in that log:

;;; (EXT:RUN-PROGRAM "gcc" ("-I." "-I/usr/include/" "-D_GNU_SOURCE" "-D_FILE_OFFSET_BITS=64" "-O2" "-flto=auto" "-ffat-lto-objects" "-fexceptions" "-g" "-grecord-gcc-switches" "-pipe" "-Wall" "-Werror=format-security" "-Wp,-D_FORTIFY_SOURCE=2" "-Wp,-D_GLIBCXX_ASSERTIONS" "-specs=/usr/lib/rpm/redhat/redhat-hardened-cc1" "-fstack-protector-strong" "-specs=/usr/lib/rpm/redhat/redhat-annobin-cc1" "-m64" "-mtune=generic" "-fasynchronous-unwind-tables" "-fstack-clash-protection" "-fcf-protection" "-Wno-unused" "-Wno-return-type" "-Wno-unknown-pragmas" "-fPIC" "-D_THREAD_SAFE" "-Dlinux" "-O2" "-c" "binary-ecl/maxima-package.c" "-o" "binary-ecl/maxima-package.o")):
;;; cc1: fatal error: inaccessible plugin file /sage/local/lib/gcc/x86_64-pc-linux-gnu/10.3.0/plugin/annobin.so expanded from short plugin name annobin: No such file or directory
;;; compilation terminated.
;      - Loading binary file "binary-ecl/maxima-package.fas" 
Condition of type: FILE-ERROR
Filesystem error with pathname #P"/sage/local/var/tmp/sage/build/maxima-5.45.0/src/src/binary-ecl/maxima-package.fas".

[Matthias Köppe]

Looks like combining system ecl with our gcc does not work because it uses Fedora-specific compiler options.

This will go away when we can accept system gcc11...

[Dima Pasechnik]

that annobin-related gcc/ecl bug was also reported on sage-devel

Redhat/Fedora? does something special with annobin, perhaps that's the fallout we see here

[Matthias Köppe]

I've opened #32093 for this

[Matthias Köppe]

Last 10 new commits:

​3c131f7	wip tarball
​11c80aa	updated tarball with PRs 377 and 378
​c9c50e3	build/pkgs/pynac: Upgrade to 0.7.29
​ee34993	Merge #31694
​228ea28	build/pkgs/ecl/spkg-configure.m4: Add SAGE_SPKG_DEPCHECK for gcc
​a22e24c	Merge #32093
​5023898	build/pkgs/lcalc/patches: Add https://github.com/cschwan/sage-on-gentoo/blob/master/sci-mathematics/lcalc/files/lcalc-1.23-gcc11.patch
​26caef1	build/pkgs/lcalc/patches: Remove the superseded Lcommon.h.patch
​6340e53	Merge #31837
​f47c47e	build/pkgs/gcc/spkg-configure.m4, .homebrew-build-env: Accept gcc 11.x

[Dima Pasechnik]

Does this work with #32037 and without lcalc patch?

[Matthias Köppe]

This is unknown. I don't think portability testing has been run on #32037.

[Dima Pasechnik]

Replying to mkoeppe:

This is unknown. I don't think portability testing has been run on #32037.

they are now running on ​https://github.com/sagemath/sagetrac-mirror/actions/runs/1009717554

[Dima Pasechnik]

gengetopt does not build on ubunty-trusty. Arrgh...

[Matthias Köppe]

With all dependencies now merged in 9.4.beta5, let's get this in please.

[Dima Pasechnik]

I believe it actually works on Fedora 34, right?

[Matthias Köppe]

Thanks, yes, I think so

[Enrique Artal Bartolo]

It works for me in Fedora 34

[Volker Braun]

I've created #33981 for latte_int