Bug #13350: File.read :newline option not respected on Linux

• Status: Closed
• Assignee: nobu (Nobuyoshi Nakada)
• Target version: -
• ruby -v: ruby 2.5.0dev (2017-03-21 trunk 58044) [x86_64-linux]
• Backport: 2.2: UNKNOWN, 2.3: UNKNOWN, 2.4: UNKNOWN

[larskanis (Lars Kanis)]

Reproducable in irb

All examples run with:
Encoding.default_external = 'utf-8' Encoding.default_internal = nil

on Linux (unexpected behavior)

Prepare a file with CRLF newline characters
File.write("crlf.txt", "a\nb\nc", newline: :crlf) # => 7 File.binread("crlf.txt") # => "a\r\nb\r\nc"

While read the files newline characters are only converted if internal and external encoding differ like here:
File.read("crlf.txt", mode: "r:utf-8:cp850", newline: :universal) # => "a\nb\nc"

But if the encoding isn't changed, the newline character is not converted:
File.read("crlf.txt", newline: :universal) # => "a\r\nb\r\nc" File.read("crlf.txt", mode: "r:utf-8:utf-8", newline: :universal) # => "a\r\nb\r\nc" File.open("crlf.txt", "r:utf-8:utf-8", newline: :universal, &:read) # => "a\r\nb\r\nc"

I would expect to get:
File.read("crlf.txt", newline: :universal) # => "a\nb\nc"

... but the :newline option seems to be ignored, when read without encoding conversion.

on Windows (expected behavior)

In contrast, on Windows the same commands returns in only 5 bytes to be written, but the content of the file is the same as on Linux:
File.write("crlf.txt", "a\nb\nc", newline: :crlf) # => 5 File.binread("crlf.txt") # => "a\r\nb\r\nc"

Due to the default CRLF->LF conversion on Windows, the newline option isn't necessary to get a CRLF->LF conversion. But setting newline to :cr shows, that the :newline option is respected as expected.
File.read("crlf.txt", newline: :universal) # => "a\nb\nc" File.read("crlf.txt", newline: :cr) # => "a\r\nb\r\nc"

[nobu (Nobuyoshi Nakada)]

newline: universal does not imply text mode.
...but it seems nice even if the newline: option enables it.

[matz (Yukihiro Matsumoto)]

It seems to be a nice idea to turn text-mode on with newline option.

Matz.

[hsbt (Hiroshi SHIBATA)]

• Status changed from Open to Assigned
• Assignee set to nobu (Nobuyoshi Nakada)

[nobu (Nobuyoshi Nakada)]

• Status changed from Assigned to Closed

Applied in changeset trunk|r59336.

---

io.c: textmode if newline decorator

• io.c (validate_enc_binmode): newline decorator implies text mode
  now. [ruby-core:80270] [Bug #13350]