Use julia-downgrade-compat@2

[JoshuaLampert]

With the new julia-downgrade-compat v2, the action switched to use Resolver.jl to find minimal compatible versions. If everything works as intended, this makes it MUCH simpler to work with and debug the downgrade action because in case of incompatible lower bounds, you don't need to resolve the incompatibilities manually by finding out a set of working minimal bounds. This should now be done by Resolver.jl.

[github-actions]

Benchmark Results (Julia v1.10)

Time benchmarks

	main	b686511...	main / b686511...
bbm_1d/bbm_1d_basic.jl - rhs!:	13.9 ± 0.32 μs	13.9 ± 0.28 μs	0.997 ± 0.031
bbm_1d/bbm_1d_fourier.jl - rhs!:	0.533 ± 0.0098 ms	0.216 ± 0.12 ms	2.46 ± 1.3
bbm_bbm_1d/bbm_bbm_1d_basic_reflecting.jl - rhs!:	0.081 ± 0.0002 ms	0.0806 ± 0.00039 ms	1 ± 0.0055
bbm_bbm_1d/bbm_bbm_1d_dg.jl - rhs!:	0.0341 ± 0.00064 ms	0.0341 ± 0.00046 ms	1 ± 0.023
bbm_bbm_1d/bbm_bbm_1d_relaxation.jl - rhs!:	27.5 ± 0.53 μs	27.8 ± 0.57 μs	0.988 ± 0.028
bbm_bbm_1d/bbm_bbm_1d_upwind_relaxation.jl - rhs!:	0.0485 ± 0.00057 ms	0.0488 ± 0.00068 ms	0.993 ± 0.018
hyperbolic_serre_green_naghdi_1d/hyperbolic_serre_green_naghdi_dingemans.jl - rhs!:	4.3 ± 0.04 μs	4.23 ± 0.03 μs	1.02 ± 0.012
kdv_1d/kdv_1d_basic.jl - rhs!:	1.42 ± 0.02 μs	1.45 ± 0.011 μs	0.979 ± 0.016
kdv_1d/kdv_1d_implicit.jl - rhs!:	1.39 ± 0.02 μs	1.45 ± 0.011 μs	0.959 ± 0.016
serre_green_naghdi_1d/serre_green_naghdi_well_balanced.jl - rhs!:	0.197 ± 0.0069 ms	0.197 ± 0.0074 ms	1 ± 0.051
svaerd_kalisch_1d/svaerd_kalisch_1d_dingemans_relaxation.jl - rhs!:	0.146 ± 0.0035 ms	0.147 ± 0.0037 ms	0.996 ± 0.035
time_to_load	2.22 ± 0.015 s	2.21 ± 0.013 s	1 ± 0.0092

Memory benchmarks

	main	b686511...	main / b686511...
bbm_1d/bbm_1d_basic.jl - rhs!:	1 allocs: 4.12 kB	1 allocs: 4.12 kB	1
bbm_1d/bbm_1d_fourier.jl - rhs!:	1 allocs: 4.12 kB	1 allocs: 4.12 kB	1
bbm_bbm_1d/bbm_bbm_1d_basic_reflecting.jl - rhs!:	5 allocs: 1.17 kB	5 allocs: 1.17 kB	1
bbm_bbm_1d/bbm_bbm_1d_dg.jl - rhs!:	10 allocs: 8.62 kB	10 allocs: 8.62 kB	1
bbm_bbm_1d/bbm_bbm_1d_relaxation.jl - rhs!:	2 allocs: 8.25 kB	2 allocs: 8.25 kB	1
bbm_bbm_1d/bbm_bbm_1d_upwind_relaxation.jl - rhs!:	2 allocs: 8.25 kB	2 allocs: 8.25 kB	1
hyperbolic_serre_green_naghdi_1d/hyperbolic_serre_green_naghdi_dingemans.jl - rhs!:	0 allocs: 0 B	0 allocs: 0 B
kdv_1d/kdv_1d_basic.jl - rhs!:	0 allocs: 0 B	0 allocs: 0 B
kdv_1d/kdv_1d_implicit.jl - rhs!:	0 allocs: 0 B	0 allocs: 0 B
serre_green_naghdi_1d/serre_green_naghdi_well_balanced.jl - rhs!:	0.075 k allocs: 0.66 MB	0.075 k allocs: 0.66 MB	1
svaerd_kalisch_1d/svaerd_kalisch_1d_dingemans_relaxation.jl - rhs!:	0.042 k allocs: 0.315 MB	0.042 k allocs: 0.315 MB	1
time_to_load	0.153 k allocs: 14.5 kB	0.153 k allocs: 14.5 kB	1

[coveralls]

Pull Request Test Coverage Report for Build 22516540989

Details

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall first build on downgrade-v2 at 98.48%

---

Totals
Change from base Build 22516477323:	98.5%
Covered Lines:	2332
Relevant Lines:	2368

---

💛 - Coveralls

[JoshuaLampert]

Hm, the Downgrade job seems to just use the same versions as the normal job (see https://github.com/NumericalMathematics/DispersiveShallowWater.jl/actions/runs/16541187956/job/46782631560?pr=227#step:8:57 and https://github.com/NumericalMathematics/DispersiveShallowWater.jl/actions/runs/16541187950/job/46782631572?pr=227#step:7:52) meaning it doesn't really test the compat bounds. Same also happens in SciML/SciMLBase.jl#1075. I think the new version is not quite ready yet.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[JoshuaLampert]

I already prepared this PR to what would need to be done once StefanKarpinski/Resolver.jl#15 is resolved.

[JoshuaLampert]

With StefanKarpinski/Resolver.jl#19 now merged, the downgrade job using julia-downgrade-compat@2 is now successful. Bumping RecursiveArrayTools.jl and StaticArrays.jl is fine I think because the previous oldest possible versions were not compatible/had bugs and were not used by the previous downgrade action version (I think because for packages with major version >0 it was less strict than the current version of the downgrade action, see also https://github.com/julia-actions/julia-downgrade-compat/blob/2f6400ecbd1a7229a6630aea5d062ac82169a396/README.md#L55). Regarding the bump of Plots.jl I wrote my analysis is in julia-actions/julia-downgrade-compat#37 (comment). I think, there is still potential for improvements in Resolver.jl, but at least the new downgrade action version seems mostly usable now for us. I'll check in a few other repos to see if there are any other issues. If not, we could try to move to the v2 of the action. What do you think, @ranocha (and thanks again for your fix in Resolver.jl)?

[ranocha]

Sounds like a good plan. Thanks!

[JoshuaLampert]

This should finally be ready.

[ranocha]

Thanks for all the work you put into fixing and updating downgrade actions!