Add Compat entries for #11140 & #11241

[ScottPJones]

This adds Compat entries for is_valid_utf32, and all of the isvalid changes.

[tkelman]

this test needs to be something like @test !isvalid(@compat Char(0x110000)) for 0.3

[ScottPJones]

Thanks, I've been trying to figure out how to make it work in 0.3

[ScottPJones]

Should be fixed now!

[tkelman]

Cool, I think this looks right. Thank you very much!

[ScottPJones]

@tkelman Can this be merged now? (so nobody has issues going back and forth between 0.4 and 0.3.x?) Thanks!

[tkelman]

Sure.

[hayd]

Not sure this was quite right as it failed travis https://travis-ci.org/JuliaLang/Compat.jl/jobs/64189228

MethodError: `isvalid` has no method matching isvalid(::Type{UTF16String}, ::UTF16String)

Does isvalid etc need to be exported outside of the if-block?

[tkelman]

Hm, that failed on 0.4.0-dev+5023, so it's actually that the test is trying to use methods that don't exist on Julia master

INFO: Testing Compat
WARNING: is_valid_utf32(str::UTF32String) is deprecated, use isvalid(str) instead.
 in depwarn at ./deprecated.jl:62
 in is_valid_utf32 at deprecated.jl:49
 in anonymous at test.jl:91
 in do_test at test.jl:49
 in include at ./boot.jl:252
 in include_from_node1 at loading.jl:133
 in process_options at ./client.jl:320
 in _start at ./client.jl:412
while loading /home/travis/.julia/v0.4/Compat/test/runtests.jl, in expression starting on line 304
ERROR: LoadError: test error in expression: isvalid(UTF16String,u16)
MethodError: `isvalid` has no method matching isvalid(::Type{UTF16String}, ::UTF16String)
Closest candidates are:
  isvalid(::Type{UTF16String}, ::AbstractArray{UInt16,N})
 in anonymous at test.jl:91
 in do_test at test.jl:49
 in anonymous at no file:320
 in include at ./boot.jl:252
 in include_from_node1 at loading.jl:133
 in process_options at ./client.jl:320
 in _start at ./client.jl:412
while loading /home/travis/.julia/v0.4/Compat/test/runtests.jl, in expression starting on line 309

[hayd]

That was the other possibility! It's annoying that this version number is not printed in the travis log, or available more cleanly from the commit without building (perhaps I'm missing a trick here?).

[tkelman]

I specifically made sure that the version number would be printed in the Travis log when we were implementing language: julia support, see https://travis-ci.org/JuliaLang/Compat.jl/jobs/64189228#L87

[hayd]

@tkelman sorry, I meant on the JuliaLang/julia travis log!

[tkelman]

I added a versioninfo() line there too: https://github.com/JuliaLang/julia/blob/b4f884d5a28579e926c97ddcb62978d1f1159217/.travis.yml#L62

[hayd]

Hmmmmmmmmm, weirdly versioninfo does show the version number locally e.g. +5016... But on travis it just has the branch name:

$ /tmp/julia/bin/julia-debug -e 'versioninfo()'
Julia Version 0.4.0-dev
...

e.g. look for "versioninfo" on this build.

[tkelman]

Oh, that is odd. I think what might be going on there is the way travis clones from github is confusing our version_git script, note the warnings a little earlier

fatal: No names found, cannot describe anything.
fatal: bad revision '^'

[ScottPJones]

Did I do something wrong? :sad:

[tkelman]

Will need to fix the method error from #90 (comment) - not sure why Travis didn't catch this the first time around

[hayd]

Travis didn't catch it cos it was Julia Version 0.4.0-dev+4850.

[ScottPJones]

OK, I figured it out... my tests were on v0.3.7, v0.3.8 and v0.4 before isvalid... and this breaks now that isvalid has been added to master, and the Compat code isn't being used, because although the Compat code handled the two argument case, the code in master missed a method for the 2-argument form...
I'm testing the fix now, but it will need to be merged into julia/master (and who knows how long that will take!) The problem is only in the test though, not in the additions to Compat.jl themselves.
That does show an issue with Compat... how do you know it gives you anything at all like actual implementation of something in the later versions of julia? You just have to be very careful... but there are no cross-version tests...

[ScottPJones]

Should be fixed by #11482 when/if that is merged...