Bug #7374: File.expand_path resolving to first file/dir instead of absolute path

• Status: Closed
• Assignee: luislavena (Luis Lavena)
• Target version: 2.0.0
• ruby -v: ruby 1.9.3p327 (2012-11-10) [i386-mingw32]

[[e-mail] (Martin Dubé)]

=begin
(({File.expand_path("./", "c:)"}))
will give "c:/$recycle.bin" as result instead of "c:/"

(({File.expand_path("./", "c:/temp)"}))
will give "c:/temp/." as result instead of "c:/temp/"

It looks like the first directory entry is returned instead of the absolute path.

=end

[[e-mail] (Martin Dubé)]

ruby 1.9.3p194 (2012-04-20) [i386-mingw32] is working fine:

irb(main):002:0> File.expand_path("./", "c:/temp")
=> "c:/temp/"

[luislavena (Luis Lavena)]

• Category set to core
• Status changed from Open to Assigned
• Assignee set to h.shirosaki (Hiroshi Shirosaki)

Martin,

Can you confirm if this happen with Ruby 2.0? (trunk, you can get automated builds from RubyInstaller CI)

Shirosaki-san,

Seems that globing has been enabled?

[luislavena (Luis Lavena)]

• Assignee changed from h.shirosaki (Hiroshi Shirosaki) to luislavena (Luis Lavena)
• Priority changed from Normal to 5
• Target version set to 2.0.0

=begin
I can confirm this happens on both 1.9.3-p327 and trunk.

Problem seems to be related to replace_to_long_name which attempts to expand the last part of the path using FindFirstFileW (globing), but is ignoring that it includes "*" in the filename.

I'm checking this and adding a test for this scenario.

Thank you for your report.
=end

[[e-mail] (Martin Dubé)]

luislavena (Luis Lavena) wrote:

Martin,

Can you confirm if this happen with Ruby 2.0? (trunk, you can get automated builds from RubyInstaller CI)

Shirosaki-san,

Seems that globing has been enabled?

Yes, same problem with r37691:

C:\ruby-2.0.0-r37691-i386-mingw32\bin>irb.bat
irb(main):001:0> File.expand_path("./*", "c:/")
=> "c:/$RECYCLE.BIN"
irb(main):002:0>

[luislavena (Luis Lavena)]

Martin,

I'm adding tests for this, do you have any other scenario that is failing and you want to report?

It will be great if we can cover all those by a single commit.

Thank you.

[[e-mail] (Martin Dubé)]

=begin
No other scenario. Caught that bug trying to install a plugin to redmine 2.1.2 with ruby 1.9.3p327

Redmine then attempt to add plugins routes with the following code:

(({Dir.glob File.expand_path("plugins/*", Rails.root) do |plugin_dir|}))

Which end up enumerating only the "." directory instead of all subdirectories in ./plugins
=end

[luislavena (Luis Lavena)]

• Status changed from Assigned to Closed
• % Done changed from 0 to 100

This issue was solved with changeset r37692.
Martin, thank you for reporting this issue.
Your contribution to Ruby is greatly appreciated.
May Ruby be with you.

---

skip long name conversion if last character is a wildcard one

• win32/file.c (replace_to_long_name): skip automatic path expansion
  when wildcard character is used. [ruby-core:49451] [Bug #7374]

• test/ruby/test_file_exhaustive.rb: add a test for above.

[luislavena (Luis Lavena)]

Martin,

I've committed some tests and a fix for this.

Automated builds will kick in soon, please check RubyInstaller CI to download binary packages.

I'll request a backport once I get confirmation this works on your end.

[h.shirosaki (Hiroshi Shirosaki)]

• Status changed from Closed to Assigned

It seems if the last part of the path contains * or ?, FindFirstFileW should not be used since FindFirstFileW expands wildcard.

Examples:
File.expand_path("foo/ab*c")
File.expand_path("foo/ab?c")

Code in file.c uses strpbrk(s, "*?") for that.
https://github.com/ruby/ruby/blob/trunk/file.c#L3141

[luislavena (Luis Lavena)]

• % Done changed from 100 to 50

h.shirosaki (Hiroshi Shirosaki) wrote:

It seems if the last part of the path contains * or ?, FindFirstFileW should not be used since FindFirstFileW expands wildcard.

Thank you Shirosaki, I did a naive fix and completely missed "?" character.

Will modify replace_to_long_name to check for this.

[luislavena (Luis Lavena)]

• Status changed from Assigned to Closed
• % Done changed from 50 to 100

This issue was solved with changeset r37700.
Martin, thank you for reporting this issue.
Your contribution to Ruby is greatly appreciated.
May Ruby be with you.

---

Skip path expansion for all wildcard characters

• win32/file.c (replace_to_long_name): skip expansion for all wildcard
  characters.
  [ruby-core:49451] [Bug #7374]

• test/ruby/test_file_exhaustive.rb: add more assertions to test.

[luislavena (Luis Lavena)]

• Status changed from Closed to Assigned
• % Done changed from 100 to 90

=begin
Sorry for the noise, but seems I've a Webrick failure caused by this that didn't catch when sent r37700 in.

test_short_filename(WEBrick::TestFileHandler) [C:/Users/Worker/Jenkins/workspace/ruby-trunk-x86-build/test/webrick/test_filehandler.rb:240]:
webrick log start:
[2012-11-17 12:49:00] INFO WEBrick 1.3.1
[2012-11-17 12:49:00] INFO ruby 2.0.0 (2012-11-17) [i386-mingw32]
[2012-11-17 12:49:00] INFO WEBrick::HTTPServer#start: pid=14444 port=62527

[2012-11-17 12:49:00] WARN the request refers nondisclosure name .htaccess'. [2012-11-17 12:49:00] ERROR /.htaccess' not found.

webrick log end.
<"404"> expected but was
<"200">.

Looking into this right now.
=end

[luislavena (Luis Lavena)]

• Status changed from Assigned to Closed
• % Done changed from 90 to 100

This issue was solved with changeset r37706.
Martin, thank you for reporting this issue.
Your contribution to Ruby is greatly appreciated.
May Ruby be with you.

---

Swap logic around wildcard character detection

• win32/file.c (replace_to_long_name): correct logic around wildcard
  characters detection and ensure wide-chars are used as pattern.
  [ruby-core:49451] [Bug #7374]

[luislavena (Luis Lavena)]

Martin, Hiroshi,

This should be finally fixed by r37706, sorry for all the inconveniences!

(note to myself: don't commit to Ruby on during night, even less on friday).