I have three suggestions concerning improving error- and warning-reporting
in 0.4.0-rc4:
(1) First, there appears to be an actual bug with error system. My program
is giving BoundsError for an array subscript out of range, but the line
number of the source statement in the error message is wrong. Even worse,
the error message is plausible because the reported line number indeed has
an array subscript operation. My program is too long (nested 'includes',
lots of packages) to helpfully post as a Julia issue, but I'm wondering if
there is some home-grown debugging using code_llvm, etc., that I could
carry out myself in order to report this issue in a more helpful way.
The other two items are not bugs but suggestions.
(2) Consider the module:
module testwarning
import Base.*
*(a::Tuple{Float64,Float64},s::Float64) = (a[1]*s,a[2]*s)
println((6.4,5.4)*3.2)
end
The first time I load this module, no problem. But every time I reload, I
get a warning about redefining *. In my actual application program, I get
screenfuls of similar warnings. My understanding is that the
software-development loop involves loading a module, executing a function,
finding bugs, and then reloading the module. So the repeated and lengthy
warnings about redefining Base operators are unhelpful in this context.
(3) Consider the following code:
immutable T
t1::Int
t2::Int
t3::Int
t4::bool
t5::Int
end
Notice the programmer's error: bool instead of Bool, a quite likely error
for a former C++ programmer. Here is the message from Julia:
ERROR: LoadError: TypeError: T: in type definition, expected Type{T}, got
Function
(etc)
while loading c:\Users\vavasis\Documents\Projects\julia\testerror1.jl, in
expression starting on line 3
The line number is misleading (it is number of the "immutable T") and the
error message itself does not mention bool. So this particular error
message could be improved.
-- Steve Vavasis
