Truncated error, with message SYSTEM: show(lasterr) caused an error

[dpsanders]

With my branch parametrised_intervals of ValidatedNumerics
(https://github.com/dpsanders/ValidatedNumerics.jl/tree/parametrised_intervals):

julia> using ValidatedNumerics

julia> a = @interval(1)
[1e+00, 1e+00] with 53 bits of precision

julia> z = zero(a)
[0e+00, 0e+00] with 53 bits of precision

julia> z == 0.0
ERROR: `BigFloat` has no method matching BigFloat(
 in promote at promotion.jl:123
 in == at promotion.jl:168SYSTEM: show(lasterr) caused an error

This happens on Julia v0.3 and v0.4.

[dpsanders]

This seems to have been due to a mixing missing convert method. But this is what I was hoping the error message would tell me...!

[ivarne]

I have tried various commits on the parametrised_intervals branch of ValidatedNumerics, but I can't reproduce. Do you have a commit with the error?

[dpsanders]

Thanks. I have a new branch truncated_error_message which shows the problem.
The only difference is that I commented out the first convert definition in interval_definition.jl.

[jballanc]

Just attempted reproducing this locally (OS X v10.10). On the truncated_error_message branch I get:

> julia -v
julia version 0.4.0-dev+2957
> julia -E 'Pkg.checkout("ValidatedNumerics", "truncated_error_message"); using ValidatedNumerics; a = @interval(0.0); z = zero(a); z == 0.0'
INFO: Checking out ValidatedNumerics truncated_error_message...
INFO: Pulling ValidatedNumerics latest truncated_error_message...
INFO: No packages to install, update or remove
ERROR: StackOverflowError:

Bisecting back through the commits between master and truncated_error_message, I found that dpsanders/ValidatedNumerics.jl@d23924 gave the following:

> julia -E "using ValidatedNumerics; a = @interval(0.0); z = zero(a); z == 0.0"
ERROR: no promotion exists for ValidatedNumerics.Interval{Base.MPFR.BigFloat} and Float64
 in promote_to_super at promotion.jl:154
 in == at promotion.jl:170
 in process_options at ./client.jl:248
 in _start at ./client.jl:403

which, if I understand correctly, is the desired error message. I'm not yet familiar enough with the type promotion system to say if the stack overflow is a bug in Julia or ValidatedNumerics, but the original issue seems to be resolved.

[garborg]

This just happened to me on the latest commit:

               _
   _       _ _(_)_     |  A fresh approach to technical computing
  (_)     | (_) (_)    |  Documentation: http://docs.julialang.org
   _ _   _| |_  __ _   |  Type "help()" for help.
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 0.4.0-dev+2993 (2015-01-30 01:32 UTC)
 _/ |\__'_|_|_|\__'_|  |  Commit 2d23b6f* (0 days old master)
|__/                   |  x86_64-apple-darwin14.1.0

julia> readline("comments/before_after_data_windows.csv")
ERROR: MethodError: `readline` has no method matching readline(::ASCIIString)SYSTEM: show(lasterr) caused an error

[jballanc]

Nice! I can reproduce the readline based version.

[ivarne]

try 
    readline("comments/before_after_data_windows.csv")
catch e
    showerror(STDOUT, e, catch_backtrace())
end

MethodError: `readline` has no method matching readline(::ASCIIString)ERROR: BoundsError: attempt to access ()
  at index [0]
 in getindex at /Users/ivarne/dev/julia/usr/lib/julia/sys.dylib
 in show_method_candidates at replutil.jl:213
 in showerror at replutil.jl:169
 in showerror at replutil.jl:81

So this is a bug in the code to show "closest" method matches when you get a MethodError. cc: @jhasse and @dhoegh.

The full backtrace isn't shown by default, because that could trigger a recursive error dumping.

[dhoegh]

I think there is two sepperate issues, I have fixed the one posted by @garborg in #9968

[tkelman]

@dpsanders is this fixed now? Could we reduce it to a test case to make sure it stays that way?