Fix the itruncalypse

[staticfloat]

I have to give credit to @nolta for a name that has kept me smiling all afternoon.

This tries to excise all mention in Base of uint(LITERAL) where LITERAL is 64 bits wide. This builds on my 32-bit VM, but I'd like a human eye to go over this and make sure I'm not doing anything horrendously stupid here.

[staticfloat]

@JeffBezanson What do you want to do about things such as these tests? We have some kind of interesting behavior here:

$ ./julia 
               _
   _       _ _(_)_     |  A fresh approach to technical computing
  (_)     | (_) (_)    |  Documentation: http://docs.julialang.org
   _ _   _| |_  __ _   |  Type "help()" for help.
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 0.4.0-dev+906 (2014-10-01 23:27 UTC)
 _/ |\__'_|_|_|\__'_|  |  sf/itruncalypse/1f5c846 (fork: 8 commits, 0 days)
|__/                   |  i686-linux-gnu

julia> -170141183460469231731687303715884105729
-170141183460469231731687303715884105729

julia> -9223372036854775809
ERROR: InexactError()

[tkelman]

This bootstraps on Win32, nice work, but fails the numbers tests as you saw.

        From worker 4:       * numbers
exception on 4: ERROR: InexactError()
while loading numbers.jl, in expression starting on line 69
ERROR: InexactError()
while loading numbers.jl, in expression starting on line 69

[JeffBezanson]

That's strange. All those tests should still pass.
Many thanks for doing this!

[staticfloat]

The InexactError from entering the literal -9223372036854775809 is coming from *(u::Int128, v::Int128), but I don't really know what to do here.

The crux of the problem is that widemul(int64(-1), int64(1)) throws an InexactError:

julia> widemul(int64(-1), int64(1))
rec_backtrace at /home/vagrant/buildbot/slave/build_ubuntu14_04-x86/build/usr/bin/../lib/libjulia.so (unknown line)
unknown function (ip: -1231114184)
jl_throw at /home/vagrant/buildbot/slave/build_ubuntu14_04-x86/build/usr/bin/../lib/libjulia.so (unknown line)
unknown function (ip: -1231113459)
widemul at int.jl:394
jlcall_widemul;39913 at  (unknown line)
jl_apply_generic at /home/vagrant/buildbot/slave/build_ubuntu14_04-x86/build/usr/bin/../lib/libjulia.so (unknown line)
unknown function (ip: -1231144368)
unknown function (ip: -1231148216)
unknown function (ip: -1231078163)
jl_toplevel_eval_in at /home/vagrant/buildbot/slave/build_ubuntu14_04-x86/build/usr/bin/../lib/libjulia.so (unknown line)
jl_f_top_eval at /home/vagrant/buildbot/slave/build_ubuntu14_04-x86/build/usr/bin/../lib/libjulia.so (unknown line)
eval_user_input at REPL.jl:54
jlcall_eval_user_input;39890 at  (unknown line)
jl_apply_generic at /home/vagrant/buildbot/slave/build_ubuntu14_04-x86/build/usr/bin/../lib/libjulia.so (unknown line)
anonymous at task.jl:96
jl_trampoline at /home/vagrant/buildbot/slave/build_ubuntu14_04-x86/build/usr/bin/../lib/libjulia.so (unknown line)
unknown function (ip: -1231115741)
julia_trampoline at /home/vagrant/buildbot/slave/build_ubuntu14_04-x86/build/usr/bin/../lib/libjulia.so (unknown line)
unknown function (ip: 134516370)
__libc_start_main at /lib/i386-linux-gnu/libc.so.6 (unknown line)
unknown function (ip: 134516419)
ERROR: InexactError()
 in widemul at int.jl:394

This is due to the multiplication of u1 and v0 here, which is mixing signed and unsigned integers in multiplication. Any ideas on how this might be fixed? Also, I wonder why these errors are appearing on 32-bit but not on 64-bit builds?

[JeffBezanson]

Reinterpreting u1 as unsigned will probably work since that's basically
what we did before.

[staticfloat]

Unfortunately I don't really have the time to figure out what to do with these functions. There is a lot of mixing uint and int in this file, basically every time you have uX * vY where X != Y. I realize it's hard to debug this without a 32-bit machine, so if you want access to my 32-bit VM I can send you an email, but I don't think I'll be able to solve this myself for a couple days.

[JeffBezanson]

Looks like you forgot to use hashis_seed here.

[staticfloat]

Thanks.

[JeffBezanson]

Not a big deal, but all the uint() calls on the constants can also be removed now.

[staticfloat]

I like saving harddrive space by cutting down on character counts just as much as anyone.