Boost upgrade from 1.74 -> 1.87

[rfaasse]

Description
The current version of boost used in the GitHub pipelines is 1.74. Since that is more than 4 years old, this issue proposes to upgrade to the latest available stable release, which is 1.87.

One of the main reasons for doing this upgrade could be that Boost 1.74 doesn't build with C++ 20. It seems boost/ublas uses the std::allocator<T>::construct function, which is deprecated in C++17 and removed in C++20. This issue is resolved in version 1.75 (see release notes. If we want to move to newer C++ standards and use the latest technology we will need to upgrade boost.

@KratosMultiphysics/technical-committee, would it be possible to look into this?

[loumalouomega]

@roigcarlo FYI

[roigcarlo]

I can change the CI no problem

Probably I will wait until #13060 Is merged if you don't mind

[matekelemen]

Do we have plans adopting C++20 (or newer)?

[loumalouomega]

Do we have plans adopting C++20 (or newer)?

Now that we are not forced to use CentOS anymore we are not that much strongly enforced to be compatible with old compilers. FYI @KratosMultiphysics/altair

[loumalouomega]

A priori we are interested in std::mdspan, but it comes with C++23.

[rfaasse]

Do we have plans adopting C++20 (or newer)?

In the GeoMechanics app we have experimented a bit with C++ 20 and there are quite some nice new features, such as concepts (especially interesting in a code base like Kratos which uses templates quite heavily), ranges (more compact forms of standard algorithms), mathematical constants, more constexpr functionality (which could improve runtime performance) and more, so in our opinion, taking this step is certainly worthwhile

[matekelemen]

@rfaasse I couldn't agree more and I'm very happy that you brought it up. Apart from coroutines, everything in C++20 would be a massive boon to kratos!

I'm very curious whether we could cut compile times by adopting C++20 modules in core.

[loumalouomega]

@rfaasse I couldn't agree more and I'm very happy that you brought it up. Apart from coroutines, everything in C++20 would be a massive boon to kratos!

I'm very curious whether we could cut compile times by adopting the C++20 modules in core.

Indeed

[RiccardoRossi]

+1 for c++20! Btw, mdspan has backports... El mar, 4 feb 2025, 16:47, Vicente Mataix Ferrándiz < ***@***.***> escribió:

…

@rfaasse <https://github.com/rfaasse?rgh-link-date=2025-02-04T15%3A44%3A52.000Z> I couldn't agree more and I'm very happy that you brought it up. Apart from coroutines, everything in C++20 would be a massive boon to kratos! I'm very curious whether we could cut compile times by adopting the C++20 modules in core. Indeed — Reply to this email directly, view it on GitHub <#13093 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB5PWEP5TNXTPW52CL2YKST2ODOH3AVCNFSM6AAAAABWOQBVK6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMZUGM3DONJVHA> . You are receiving this because you are on a team that was mentioned.Message ID: ***@***.***>

[loumalouomega]

Yes I know, the original implementation is the kokkos team. El mar., 4 feb. 2025 17:18, Riccardo Rossi ***@***.***> escribió:

…

+1 for c++20! Btw, mdspan has backports... El mar, 4 feb 2025, 16:47, Vicente Mataix Ferrándiz < ***@***.***> escribió: > @rfaasse > <https://github.com/rfaasse?rgh-link-date=2025-02-04T15%3A44%3A52.000Z> I > couldn't agree more and I'm very happy that you brought it up. Apart from > coroutines, everything in C++20 would be a massive boon to kratos! > > I'm very curious whether we could cut compile times by adopting the C++20 > modules in core. > > Indeed > > — > Reply to this email directly, view it on GitHub > < #13093 (comment)>, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/AB5PWEP5TNXTPW52CL2YKST2ODOH3AVCNFSM6AAAAABWOQBVK6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMZUGM3DONJVHA> > . > You are receiving this because you are on a team that was mentioned.Message > ID: ***@***.***> > — Reply to this email directly, view it on GitHub <#13093 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEYQZAH6HH3TH3IF4MYGRW32ODR4FAVCNFSM6AAAAABWOQBVK6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMZUGQ2TEOBXGE> . You are receiving this because you commented.Message ID: ***@***.***>

[RiccardoRossi]

this was discussed in @KratosMultiphysics/technical-committee .

We agree about bumping up Boost, nevertheless we are still not ready for bumping up the official version to c++20.
We will keep this proposal in our todo list and verify periodically when we can do this.