[Dune-devel] First candidate for upcoming release 2.8.0 available for testing

Fri Aug 6 17:44:49 CEST 2021

Hi Christoph, Hi Markus,

thanks a lot for organizing this first release candidate.

I tried out my stack for dune-copasi with 5 settings and here is my feedback:

- Dependencies: common, geometry, grid, uggrid, istl, localfunctions, functions, logging, multidomaingrid, typetree, testtools, pdelab (custom branch).

- MPI is disabled

- QuadMath is disabled

- Using GMP

- C++ Flags: visibility hidden and position independent code

- Using python venv for testtools (except on Windows)

- Build type: Release

- Every module is installed and its build dir is removed before the next one is processed: This ensures that headers are correctly installed on every single module.

- CMake Warnings (same for all systems except Debian):

  - (dev) Warnign Policy not set: CMP0087

  - (dev) Warning at cmake/modules/DuneGridMacros.cmake:20: include(Psurface)
- CMake Error (same for all systems):

   - CMake module does not exist DuneCxaDemangle. (Easy fix both in source and in dune-common, see MR !987)

Setup 1:

- OS: Ubuntu 16.04.7 LTS

- Compiler: GNU 9.3.0

- Using dune as shared lib

- C++ Warnings:
   - UGGrid:

     - /tmp/setup-dune/dune-uggrid/dune/uggrid/gm/ugio.cc:3360:39: warning: argument 1 value ‘18446744073709551614’ exceeds maximum object size 9223372036854775807 [-Walloc-size-larger-than=]
     - /tmp/setup-dune/dune-uggrid/dune/uggrid/gm/er.cc:1664:50: warning: iteration 4 invokes undefined behavior [-Waggressive-loop-optimizations]
     - /tmp/setup-dune/dune-uggrid/dune/uggrid/gm/er.cc:1666:40: warning: iteration 4 invokes undefined behavior [-Waggressive-loop-optimizations]

     - /tmp/setup-dune/dune-uggrid/dune/uggrid/gm/ugio.cc:3360:39: warning: argument 1 value ‘18446744073709551614’ exceeds maximum object size 9223372036854775807 [-Walloc-size-larger-than=]

Setup 2:

- OS: MacOSX 10.15.7

- Compiler: AppleClang 12.0.0.12000032

- Using dune as shared lib

- C++ Warnings:

   - UGGrid:

     - 13 warnings at /tmp/setup-dune/dune-uggrid/dune/uggrid/domain/std_domain.cc: array index XX is past the end of the array (which contains YY element) [-Warray-bounds]
        - Source lines: 1886:25-2086:5-2088:5-2089:5-2099:25-2099:51-2100:17-2108:25-2108:51-2108:61-2136:29-2136:60-2137:21
     - /tmp/setup-dune/dune-uggrid/dune/uggrid/gm/ugm.cc:191:11: warning: result of comparison of constant 32 with expression of type 'UG::D2::GM_OBJECTS' is always false [-Wtautological-constant-out-of-range-compare]

Setup 3 and 4:
- OS: Microsoft Windows Server 2019 10.0.17763
- Compiler: mingw-w64-x86_64-gcc / GNU 10.3.0 and mingw-w64-i686-gcc / GNU 10.3.0 (this is for 32bit Windows)
- Using dune as static lib
- C++ Warnings:
   - UGGrid: Same warnings as for Ubuntu
- CMake Warning at C:/smelibs/share/dune/cmake/modules/DuneSymlinkOrCopy.cmake:196 (message): Your module wanted to create symlinks, but you cannot do that on your platform.

Setup 5:

- Base Docker Image: docker/ci/debian:10

- Compiler: GNU 8.3.0

- Using dune as shared lib

- C++ Warnings:
   - UGGrid: Same warnings as for Ubuntu

So I have same remarks as Simon regarding uggrid. But in my experience uggrid has always being a source of many compiler warnings and is not particular to this release.

Best Regards and Thanks!

Santiago

________________________________
From: Dune-devel <dune-devel-bounces at lists.dune-project.org> on behalf of Christoph Grüninger <foss at grueninger.de>
Sent: Thursday, August 5, 2021 12:26:06 AM
To: dune-devel at lists.dune-project.org
Subject: Re: [Dune-devel] First candidate for upcoming release 2.8.0 available for testing

Hi Simon,
thanks for testing and providing feedback!

> cmake warnings:
> - Policy CMP0087 is not set (I think we should fix this in the next
> release)

CMP0087 was introduced in CMake 3.14, we require 3.13. I have not found
a way to silence this warning, beside explicitly setting the policy in
every module's root CMakeLists.txt.

> - WARNING: could not find module 'dune-python', [...] 'dune-python' is
> suggested by dune-foamgrid

Should be fixed within dune-foamgrid. Not sure whether it is enough to
just drop the suggestion. Maybe the Python code has to be adjusted, too.

> - dune-common/dune/common/reservedvector.hh:70:5: note: variable
> tracking size limit exceeded with ‘-fvar-tracking-assignments’, retrying
> without (this I get a lot, but don't know what to do about it)

GCC gives up on tracking. You can increase the buffer size by passing
--param=max-vartrack-size= with the right number. This will increase the
compile time and memory consumption.
Or we could simplify our code here, maybe with never language features.
Not sure if it is worth the effort or even possible.

> So, maybe we should fix the onedgrid errors that occur with -std=c++20.
> It is easy to fix. There just the template parameters on the constructor
> and destructor must be removed.

Can you please create an MR?

> Most of the warnings come from the dune-uggrid module. There is still
> a lot of cleanup necessary. This is not a quick fix.

Low hanging fruits were already collected. Oliver usually welcomes any
clean-up work for UG.

Bye
Christoph

_______________________________________________
Dune-devel mailing list
Dune-devel at lists.dune-project.org
https://lists.dune-project.org/mailman/listinfo/dune-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.dune-project.org/pipermail/dune-devel/attachments/20210806/921d6167/attachment.htm>