fixes for 0.7 #671

fredrikekre · 2018-07-05T07:18:27Z

This is #623 + #667 + two additional fixes. Closes #661, #654, #637.

This PR works locally on both 0.7 and 0.6 for me, would be great to get IJulia to work properly on 0.7!

Julia 0.7 changed the `String()` constructor to deplete `Vector{UInt8}` arguments, so we must `copy()` them first to avoid clobbering.

add build.log to .gitignore

stevengj · 2018-07-05T15:23:35Z

src/inline.jl

@@ -1,6 +1,9 @@
 import Base: display, redisplay

-immutable InlineDisplay <: Display end
+@static if !isdefined(Base, :AbstractDisplay) # remove when Compat.jl#482 is merged and tagged


Shouldn't this be:

if !isdefined(Base, :AbstractDisplay) && !isdefined(Compat, :AbstractDisplay) ...

(No need for @static in global scope unless the expressions have ccall or macros...)

Actually, this if statement can be deleted completely as the AbstractDisplay definition was tagged in Compat v0.52.0

stevengj · 2018-07-05T15:24:57Z

Great, thanks for working on this! Sorry I haven't had time to bang on it.

Do things like println now work in both 0.6 and 0.7?

fredrikekre · 2018-07-05T16:47:03Z

Pushed one more deprecation fix and updated to use Compat.AbstractDisplay directly.

Do things like println now work in both 0.6 and 0.7?

Yes! 🎉

fredrikekre · 2018-07-06T07:09:29Z

src/stdio.jl

+            logger = Base.CoreLogging._global_logstate.logger
+
+            # Override logging if it's pointing at the default stderr
+            if logger.stream == Base.stderr


Actually, there is something bugged here; this is only true on the first call, don't we need to reset it somewhere @staticfloat ? On this branch; with a @show logger.stream == Base.stderr inserted:

julia> using IJulia julia> mktemp() do path, io redirect_stderr(IJulia.IJuliaStdio(io, "stderr")) do @warn("warn") end flush(io) seek(io, 0) @show read(io, String) end logger.stream == Base.stderr = true read(io, String) = "\e[33m\e[1m┌ \e[22m\e[39m\e[33m\e[1mWarning: \e[22m\e[39mwarn\n\e[33m\e[1m└ \e[22m\e[39m\e[90m@ Main REPL[2]:3\e[39m\n" "\e[33m\e[1m┌ \e[22m\e[39m\e[33m\e[1mWarning: \e[22m\e[39mwarn\n\e[33m\e[1m└ \e[22m\e[39m\e[90m@ Main REPL[2]:3\e[39m\n" julia> mktemp() do path, io redirect_stderr(IJulia.IJuliaStdio(io, "stderr")) do @warn("warn") end flush(io) seek(io, 0) @show read(io, String) end logger.stream == Base.stderr = false read(io, String) = ""

Perhaps it does not matter; this seems to be called only once per session anyways and it works in a notebook

Would it not be simpler to just replace the global logger instead of checking the internal state?

fredrikekre · 2018-07-06T09:39:31Z

I updated the logging redirect, now things seem to work:

stevengj · 2018-07-06T13:17:01Z

src/handlers.jl

@@ -138,7 +138,7 @@ function complete_request(socket, msg)

    codestart = find_parsestart(code, cursorpos)
    comps, positions = Base.REPLCompletions.completions(code[codestart:end], cursorpos-codestart+1)


Probably need:

if VERSION < v"0.7.0-DEV.3500" # julia#25544 import Base: REPLCompletions else import REPL: REPLCompletions end

above so that we can just call this as REPLCompletions.completions with no deprecation warning.

Thanks, there was a similar if VERSION check at the top so I added the imports there.

fredrikekre · 2018-07-06T13:34:55Z

While trying the tab-completion for deprecations I got this

BoundsError: attempt to access 0-element Array{Int64,1} at index [0]

Seems to be a bug here

IJulia.jl/src/handlers.jl

Lines 144 to 145 in 90b0165

    
           elseif isempty(positions) # true if comps to be inserted without replacement 
        
               cursor_start = (cursor_end = ind2chr(msg, code, last(positions)))

where we try to extract the last element although we just made sure it's empty.

stevengj · 2018-07-06T13:42:21Z

Maybe should be cursor_start = cursor_end = cursor_chr, i.e. just merge it with the previous if clause to make it if isempty(comps) || isempty(positions)

(Or maybe we should use first(positions) instead of last? But probably there isn't any case where tab completion with an empty positions range wants to insert text at someplace other than the cursor location?)

fredrikekre · 2018-07-06T13:48:35Z

src/handlers.jl

-    comps, positions = Base.REPLCompletions.completions(code[codestart:end], cursorpos-codestart+1)
-    positions += codestart-1
+    comps, positions = REPLCompletions.completions(code[codestart:end], cursorpos-codestart+1)
+    positions = positions .+ codestart-1


Nevermind, it is this fix that is missing a . on the - and the range turned into an array. Of course last works for an empty range.

fredrikekre · 2018-07-06T13:56:39Z

src/handlers.jl

-    comps, positions = Base.REPLCompletions.completions(code[codestart:end], cursorpos-codestart+1)
-    positions += codestart-1
+    comps, positions = REPLCompletions.completions(code[codestart:end], cursorpos-codestart+1)
+    positions = positions .+ codestart .- 1


Hmm, still not correct; this returns a Vector on Julia 0.6.

- warn() -> at-warn - emtpy try-catch - remove_destination -> force - start -> firstindex - endof -> lastindex - broadcast +

stevengj · 2018-07-06T14:47:06Z

LGTM, ready to merge?

stevengj · 2018-07-06T14:48:16Z

src/IJulia.jl

@@ -198,7 +202,7 @@ function clear_history(indices)
 end

 # since a range could be huge, intersect it with 1:n first
-clear_history{T<:Integer}(r::AbstractRange{T}) =
+clear_history(r::AbstractRange{T}) where {T<:Integer} =


Can just be clear_history(r::AbstractRange{<:Integer}) since we don't use T.

stevengj · 2018-07-06T14:49:23Z

src/execute_request.jl

@@ -58,30 +56,50 @@ const displayqueue = Any[]

 # remove x from the display queue
 function undisplay(x)
-    i = findfirst(equalto(x), displayqueue)
-    if i > 0
+    i = findfirst(isequal(x), displayqueue)


Shouldn't this be Compat.findfirst to get the version that returns nothing on 0.6? Then you can drop the i > 0 check.

stevengj · 2018-07-06T14:50:37Z

src/execute_request.jl

 function show_bt(io::IO, top_func::Symbol, t, set)
    # follow PR #17570 code in removing top_func from backtrace
-    eval_ind = findlast(addr->Base.REPL.ip_matches_func(addr, top_func), t)
-    eval_ind != 0 && (t = t[1:eval_ind-1])
+    eval_ind = findlast(addr->ip_matches_func(addr, top_func), t)


Similar to above, use Compat.findlast and then you can skip the eval_ind != 0 check.

stevengj · 2018-07-06T16:47:39Z

I'm just going to merge this as-is so that people can use IJulia with 0.7, and then add more fixes in another PR.

fredrikekre · 2018-07-06T21:32:09Z

Thanks! I have to admit that I did not review the changes from #623 myself.

stevengj and others added 15 commits June 7, 2018 14:14

0.7 compatibility (mostly)

507db29

update appveyor to test nightly too

f871e6a

fix merge error

adc6269

missing compat in kernel.jl

d1d65bb

REPL.REPLCompletions is now in the stdlib

1d05a2c

don't force Compat into user notebooks

9bba2a6

fix more deprecations and a crash in heartbeat

bbdebe9

fix more 0.7 deprecations

b25a960

don't use uppercase STDIO on 0.7

a82da39

fix merge conflict

d17896a

whitespace

46ea1b5

Add logging override for 0.7

0338528

Fix stdout messages not being sent

23e5178

Julia 0.7 changed the `String()` constructor to deplete `Vector{UInt8}` arguments, so we must `copy()` them first to avoid clobbering.

fix empty try-catch deprecation

cbe06f3

add build.log to .gitignore

add Conda before using it

10ddc7c

fredrikekre force-pushed the fe/nightly branch from 4726105 to 10ddc7c Compare July 5, 2018 07:53

stevengj reviewed Jul 5, 2018

View reviewed changes

fredrikekre added 2 commits July 5, 2018 18:41

fix deprecation from spawn

543305e

use Compat.AbstractDisplay directly

6e73373

fredrikekre commented Jul 6, 2018

View reviewed changes

stevengj reviewed Jul 6, 2018

View reviewed changes

This was referenced Jul 6, 2018

0.7 compatibility (mostly) #623

Closed

Add logging override for 0.7 #667

Closed

fredrikekre force-pushed the fe/nightly branch from f526e35 to b444612 Compare July 6, 2018 13:31

fredrikekre commented Jul 6, 2018

View reviewed changes

fredrikekre force-pushed the fe/nightly branch from b444612 to d1caa23 Compare July 6, 2018 13:49

fredrikekre commented Jul 6, 2018

View reviewed changes

fredrikekre force-pushed the fe/nightly branch from d1caa23 to 57524f9 Compare July 6, 2018 14:00

fredrikekre added 3 commits July 6, 2018 16:02

Misc 0.7 changes

beb1cf3

- warn() -> at-warn - emtpy try-catch - remove_destination -> force - start -> firstindex - endof -> lastindex - broadcast +

replace global logger in init()

fd0beb4

test explicit stdout/stderr printing

0d07dbd

fredrikekre force-pushed the fe/nightly branch from 57524f9 to 0d07dbd Compare July 6, 2018 14:02

stevengj reviewed Jul 6, 2018

View reviewed changes

stevengj merged commit 533ca5d into master Jul 6, 2018

stevengj deleted the fe/nightly branch July 6, 2018 16:47

stevengj mentioned this pull request Jul 6, 2018

a few more 0.7 updates #672

Merged

cstjean mentioned this pull request Sep 24, 2018

Compat.stderr != STDERR #654

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fixes for 0.7 #671

fixes for 0.7 #671

fredrikekre commented Jul 5, 2018 •

edited by stevengj

Loading

stevengj Jul 5, 2018

stevengj Jul 5, 2018

stevengj commented Jul 5, 2018

fredrikekre commented Jul 5, 2018

fredrikekre Jul 6, 2018

fredrikekre Jul 6, 2018

fredrikekre Jul 6, 2018

fredrikekre commented Jul 6, 2018

stevengj Jul 6, 2018

fredrikekre Jul 6, 2018

fredrikekre commented Jul 6, 2018

stevengj commented Jul 6, 2018 •

edited

Loading

fredrikekre Jul 6, 2018

fredrikekre Jul 6, 2018

stevengj commented Jul 6, 2018

stevengj Jul 6, 2018

stevengj Jul 6, 2018 •

edited

Loading

stevengj Jul 6, 2018

stevengj commented Jul 6, 2018

fredrikekre commented Jul 6, 2018

		@@ -138,7 +138,7 @@ function complete_request(socket, msg)

		codestart = find_parsestart(code, cursorpos)
		comps, positions = Base.REPLCompletions.completions(code[codestart:end], cursorpos-codestart+1)

fixes for 0.7 #671

fixes for 0.7 #671

Conversation

fredrikekre commented Jul 5, 2018 • edited by stevengj Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

stevengj commented Jul 5, 2018

fredrikekre commented Jul 5, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fredrikekre commented Jul 6, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fredrikekre commented Jul 6, 2018

stevengj commented Jul 6, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

stevengj commented Jul 6, 2018

Choose a reason for hiding this comment

stevengj Jul 6, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

stevengj commented Jul 6, 2018

fredrikekre commented Jul 6, 2018

fredrikekre commented Jul 5, 2018 •

edited by stevengj

Loading

stevengj commented Jul 6, 2018 •

edited

Loading

stevengj Jul 6, 2018 •

edited

Loading