colonful reshape may spuriously throw DivideError

[nsajko]

julia> reshape([], (1, :)) == reshape([], (1, 0))
true

julia> reshape([], (1, :)) == reshape([], (1, 0)) == reshape([], (0, :))  # shouldn't throw
ERROR: DivideError: integer division error
Stacktrace:
 [1] div
   @ ./int.jl:295 [inlined]
 [2] divrem
   @ ./div.jl:203 [inlined]
 [3] divrem
   @ ./div.jl:181 [inlined]
 [4] _reshape_uncolon
   @ ./reshapedarray.jl:138 [inlined]
 [5] reshape(parent::Vector{Any}, dims::Tuple{Int64, Colon})
   @ Base ./reshapedarray.jl:129
 [6] top-level scope
   @ REPL[2]:1

julia> reshape([1], (0, :))  # should throw, but not `DivideError`
ERROR: DivideError: integer division error
Stacktrace:
 [1] div
   @ ./int.jl:295 [inlined]
 [2] divrem
   @ ./div.jl:203 [inlined]
 [3] divrem
   @ ./div.jl:181 [inlined]
 [4] _reshape_uncolon
   @ ./reshapedarray.jl:138 [inlined]
 [5] reshape(parent::Vector{Int64}, dims::Tuple{Int64, Colon})
   @ Base ./reshapedarray.jl:129
 [6] top-level scope
   @ REPL[2]:1

julia> b = prevpow(2, typemax(Int))
4611686018427387904

julia> iszero(b*b)  # `b` is a square root of zero, thus a zero divisor
true

julia> reshape([1], (b, :, b))  # should throw, but not `DivideError`
ERROR: DivideError: integer division error
Stacktrace:
 [1] div
   @ ./int.jl:295 [inlined]
 [2] divrem
   @ ./div.jl:203 [inlined]
 [3] divrem
   @ ./div.jl:181 [inlined]
 [4] _reshape_uncolon
   @ ./reshapedarray.jl:138 [inlined]
 [5] reshape(parent::Vector{Int64}, dims::Tuple{Int64, Colon, Int64})
   @ Base ./reshapedarray.jl:129
 [6] top-level scope
   @ REPL[6]:1

julia> reshape([], (b, :, b))
ERROR: DivideError: integer division error
Stacktrace:
 [1] div
   @ ./int.jl:295 [inlined]
 [2] divrem
   @ ./div.jl:203 [inlined]
 [3] divrem
   @ ./div.jl:181 [inlined]
 [4] _reshape_uncolon
   @ ./reshapedarray.jl:138 [inlined]
 [5] reshape(parent::Vector{Any}, dims::Tuple{Int64, Colon, Int64})
   @ Base ./reshapedarray.jl:129
 [6] top-level scope
   @ REPL[4]:1

For the last example, the expected result depends on the resolution of #54244, but in any case it shouldn't throw DivideError.

I've got a fix, but waiting for a decision on #54244.

julia> versioninfo()
Julia Version 1.12.0-DEV.387
Commit b5bfd83a3d0 (2024-04-22 13:12 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 8 × AMD Ryzen 3 5300U with Radeon Graphics
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, znver2)
Threads: 1 default, 0 interactive, 1 GC (on 8 virtual cores)
Environment:
  JULIA_NUM_PRECOMPILE_TASKS = 3
  JULIA_PKG_PRECOMPILE_AUTO = 0

This is the affected code:

julia/base/reshapedarray.jl

Lines 130 to 141 in 8f6418e

	@inline function _reshape_uncolon(A, dims)
	@noinline throw1(dims) = throw(DimensionMismatch(string("new dimensions $(dims) ",
	"may have at most one omitted dimension specified by `Colon()`")))
	@noinline throw2(A, dims) = throw(DimensionMismatch(string("array size $(length(A)) ",
	"must be divisible by the product of the new dimensions $dims")))
	pre = _before_colon(dims...)
	post = _after_colon(dims...)
	_any_colon(post...) && throw1(dims)
	sz, remainder = divrem(length(A), prod(pre)*prod(post))
	remainder == 0 || throw2(A, dims)
	(pre..., Int(sz), post...)
	end