Check CI on Julia v1.9.2 and v1.9.3

[ranocha]

No description provided.

[codecov]

Codecov Report

Patch coverage has no change and project coverage change: -13.69% ⚠️

Comparison is base (e22b11e) 96.14% compared to head (e7b5abc) 82.45%.

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #1618       +/-   ##
===========================================
- Coverage   96.14%   82.45%   -13.69%     
===========================================
  Files         414      414               
  Lines       33942    33900       -42     
===========================================
- Hits        32633    27951     -4682     
- Misses       1309     5949     +4640     

Flag	Coverage Δ
unittests	82.45% <ø> (-13.69%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

see 99 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ranocha]

I can reproduce the allocation test failures

• https://github.com/trixi-framework/Trixi.jl/actions/runs/6020979393/job/16333094917?pr=1618#step:7:8348
• https://github.com/trixi-framework/Trixi.jl/actions/runs/6020979393/job/16333095306?pr=1618

locally only with julia --check-bounds=yes - not with default settings

[ranocha]

Reduced example:

Julia v1.9.3 default:

julia> struct BroadcastRef end

julia> Base.broadcastable(b::BroadcastRef) = Ref(b)

julia> struct BroadcastTuple end

julia> Base.broadcastable(b::BroadcastTuple) = (b,)

julia> foo(a, _) = a
foo (generic function with 1 method)

julia> function test(arg)
           n = 4
           dst = rand(n, n)
           src = rand(n, n)
           @allocated begin
               @. dst = foo(src, arg)
           end
       end
test (generic function with 1 method)

julia> test(BroadcastRef()), test(BroadcastTuple())
(0, 0)

Julia v1.9.3 with --check-bounds=yes:

julia> struct BroadcastRef end

julia> Base.broadcastable(b::BroadcastRef) = Ref(b)

julia> struct BroadcastTuple end

julia> Base.broadcastable(b::BroadcastTuple) = (b,)

julia> foo(a, _) = a
foo (generic function with 1 method)

julia> function test(arg)
           n = 4
           dst = rand(n, n)
           src = rand(n, n)
           @allocated begin
               @. dst = foo(src, arg)
           end
       end
test (generic function with 1 method)

julia> test(BroadcastRef()), test(BroadcastTuple())
(8, 0)

[sloede]

Wow! That sounds like a non-trivial MWE to create! I just confirmed:

Julia v1.9.2 with --check-bounds=yes:

julia> struct BroadcastRef end

julia> Base.broadcastable(b::BroadcastRef) = Ref(b)

julia> struct BroadcastTuple end

julia> Base.broadcastable(b::BroadcastTuple) = (b,)

julia> foo(a, _) = a
foo (generic function with 1 method)

julia> function test(arg)
           n = 4
           dst = rand(n, n)
           src = rand(n, n)
           @allocated begin
               @. dst = foo(src, arg)
           end
       end
test (generic function with 1 method)

julia> test(BroadcastRef()), test(BroadcastTuple())
(0, 0)

This confirms that is a regression since v1.9.2. If you report it, can you ping me please?

[ranocha]

Wow! That sounds like a non-trivial MWE to create!

Oh yes... it took me quite some time to get there 🙈

[DanielDoehring]

Interesting that our tests seem to catch more than the ones of the original Julia projects 😳