Bug #5297: Either File.expand_path or File.join is corrupting string encoding

• Status: Closed
• Assignee: usa (Usaku NAKAMURA)
• Target version: 2.0.0
• ruby -v: ruby 1.9.4dev (2011-09-07 trunk 33212) [i386-mingw32]

[luislavena (Luis Lavena)]

Hello,

While working on some API improvements for Windows, found the following issue:

https://gist.github.com/1202366

V:\fóñè>ruby -v
ruby 1.9.4dev (2011-09-07 trunk 33212) [i386-mingw32]

V:\fóñè>chcp 1252
Active code page: 1252

V:\fóñè>ruby -e "puts Encoding.default_external"
Windows-1252

V:\fóñè>irb
irb(main):001:0> a = File.expand_path "."
=> "V:/fóñè"
irb(main):002:0> a.encoding
=> #
irb(main):003:0> b = Dir.glob("../*").first
=> "../fóñè"
irb(main):004:0> b.encoding
=> #
irb(main):005:0> File.expand_path b
=> "V:/fóñè"
irb(main):006:0> c = File.expand_path b
=> "V:/fóñè"
irb(main):007:0> c.encoding
=> #
irb(main):008:0> d = File.join(a, "foo")
=> "V:/f\xF3\xF1\xE8/foo"
irb(main):009:0> d.encoding
=> #                          #  e = "#{a}/foo"
=> "V:/fóñè/foo"
irb(main):011:0> e.encoding
=> #
irb(main):012:0> File.open(d, "w+") { |f| f.puts "hi" }
Errno::ENOENT: No such file or directory - V:/fóñè/foo       # '
irb(main):013:0> File.open(e, "w+") { |f| f.puts "hi" }
Errno::ENOENT: No such file or directory - V:/fóñè/foo       # '
irb(main):014:0>

It is not clear why while File.expand_path worked, File.join broke but string interpolation didn't.

Even worse is that File.open failed.

I'm working on a replacement function for expand_path that rely on MultiByteToWideChar + GetFullPathNameW + WideCharToMultiByte and then uses rb_filesystem_str_new_cstr to return the string.

The funny fact is that replacement work properly:

C:\Users\Luis\Projects\oss\me\fenix>ripl -Ilib
>> require "fenix"
=> true
>> Dir.chdir "V:"
=> 0
>> Dir.pwd
=> "V:/fóñè"
>> c = Fenix::File.expand_path "."
=> "V:/fóñè"
>> c.encoding
=> #
>> File.join(c, "foo").encoding
=> #
>> d = "#{c}/foo"
=> "V:/fóñè/foo"
>> d.encoding
=> #
>> File.open(d, "w") { |f| f.puts "hi" }
=> nil

[luislavena (Luis Lavena)]

• Status changed from Open to Closed

This has been solved already associated to another bug report.

[patrickb (Patrick Bennett)]

Which other issue is this associated with?
Is this going to be patched back to 1.9.3? It's still present in 1.9.3p125

[luislavena (Luis Lavena)]

Patrick Bennett wrote:

Which other issue is this associated with?
Is this going to be patched back to 1.9.3? It's still present in 1.9.3p125

Sorry, but with released patchlevel 125 I can no longer reproduce this:

V:\fóñè>ruby -v
ruby 1.9.3p125 (2012-02-16) [i386-mingw32]

V:\fóñè>date /T
29/02/2012

V:\fóñè>time /T
02:46 p.m.

V:\fóñè>chcp
Active code page: 1252

V:\fóñè>ruby -e "puts Encoding.default_external"
Windows-1252

V:\fóñè>irb
irb(main):001:0> a = File.expand_path "."
=> "V:/fóñè"
irb(main):002:0> a.encoding
=> #
irb(main):003:0> b = Dir.glob("../*")[1]
=> "../fóñè"
irb(main):004:0> b.encoding
=> #
irb(main):005:0> c = File.expand_path b
=> "V:/fóñè"
irb(main):006:0> c.encoding
=> #
irb(main):007:0> d = File.join(a, "foo")
=> "V:/fóñè/foo"
irb(main):008:0> d.encoding
=> #
irb(main):009:0> e = "#{a}/foo"
=> "V:/fóñè/foo"
irb(main):010:0> e.encoding
=> #
irb(main):011:0> File.open(d, "w+") { |f| f.puts "hi" }
=> nil
irb(main):012:0> File.open(e, "w+") { |f| f.puts "hi" }
=> nil
irb(main):013:0> exit

[patrickb (Patrick Bennett)]

With ruby 1.9.3p125 (2012-02-16) [i386-mingw32] File.join always converts to ASCII-8BIT for me no matter the encoding passed to it.
So, using your irb example up through the File.join
irb(main):001:0> a = File.expand_path "."
=> "d:/test-streams"
irb(main):002:0> a.encoding
=> #Encoding:Windows-1252
irb(main):003:0> b = Dir.glob("../*")[1]
=> "../2dot4DSTree.reg"
irb(main):004:0> b.encoding
=> #Encoding:IBM437
irb(main):005:0> c = File.expand_path b
=> "d:/2dot4DSTree.reg"
irb(main):006:0> c.encoding
=> #Encoding:Windows-1252
irb(main):007:0> d = File.join(a, "foo")
=> "d:/test-streams/foo"
irb(main):008:0> d.encoding
=> #Encoding:ASCII-8BIT
irb(main):009:0> File.join('foo','bar').encoding
=> #Encoding:ASCII-8BIT

The result is the same regardless of my default external codepage - if I change it to 1252 as you have it then b's encoding returns as 1252 instead of 437 (my default) but File.join still returns as ascii-8bit. The fact that we;re apparently using the same ruby version is a little troubling though.

[luislavena (Luis Lavena)]

Patrick Bennett wrote:

With ruby 1.9.3p125 (2012-02-16) [i386-mingw32] File.join always converts to ASCII-8BIT for me no matter the encoding passed to it.

=> #Encoding:ASCII-8BIT

The result is the same regardless of my default external codepage - if I change it to 1252 as you have it then b's encoding returns as 1252 instead of 437 (my default) but File.join still returns as ascii-8bit. The fact that we;re apparently using the same ruby version is a little troubling though.

The problem is your system encoding.

For some reason from IBM437 to Windows-1252 on Dir.glob is not working.

Please open a separate issue.

The issue described here is about File.join messing with encoding and causing File.open to fail.