Display of exception stacks could be better

[fredrikekre]

To me the new exception stack printing (#30900) is not so clear, compare with e.g. this reduced example:

$ cat foo.jl 
function foo()
    error("hello")
end
foo()

which displays on master as

$ julia-master --color=yes foo.jl 
ERROR: Error while loading expression starting at /home/fredrik/foo.jl:4
caused by [exception 1]
hello
Stacktrace:
 [1] error(::String) at ./error.jl:33
 [...]
 [8] _start() at ./client.jl:476

and on Julia 1 as

$ julia11 --color=yes foo.jl 
ERROR: LoadError: hello
Stacktrace:
 [1] error(::String) at ./error.jl:33
 [...]
 [8] _start() at ./client.jl:436
in expression starting at /home/fredrik/foo.jl:4

I have two problems with the new printing:

1. From the output on master it took me quite a while to understand that the hello output actually came from the ERROR: two lines above. IMO it was more clear before what the error was, and I suggest this should be printed as

   $ julia-master --color=yes foo.jl 
   ERROR: hello
   while loading expression starting at /home/fredrik/foo.jl:4
   caused by [exception 1]
   Stacktrace:
    [1] error(::String) at ./error.jl:33
    [...]
    [8] _start() at ./client.jl:476
   

   instead. It seems more important to print the message before the location.

2. When seeing caused by [exception 1] I would expect a list of exceptions somewhere where I can look up what exception 1 is. What is exception 1 in the case above? It is not clear to me.

[JeffBezanson]

What's happening is that there are two errors; the original error("hello") which is the root cause, and a LoadError thrown because of that. So what it's trying to say is "a LoadError happened at this file location, caused by the error hello". So hello needs to somehow be after "caused by". That allows it to chain sensibly in the general case, e.g.

julia> try
       try
       error(1)
       catch
       error(2)
       end
       catch
       error(3)
       end
ERROR: 3
Stacktrace:
 [1] error(::Int64) at ./error.jl:42
 [2] top-level scope at REPL[10]:8
caused by [exception 2]
2
Stacktrace:
 [1] error(::Int64) at ./error.jl:42
 [2] top-level scope at REPL[10]:5
caused by [exception 1]
1
Stacktrace:
 [1] error(::Int64) at ./error.jl:42
 [2] top-level scope at REPL[10]:3

I agree [exception 1] is not too helpful though.

One possibility is to just remove LoadError, since I believe all the info should be in the stack trace now. Part of the problem is that it's not intuitive that a nested exception would be happening here.