Project

General

Profile

Bug #11363

Fix tests for String#crypt so they pass on OpenBSD

Added by jeremyevans0 (Jeremy Evans) almost 5 years ago. Updated about 1 year ago.

Status:
Closed
Priority:
Normal
Assignee:
-
Target version:
-
ruby -v:
ruby 2.3.0dev (2015-07-16 openbsd 51261) [x86_64-openbsd]
[ruby-core:70015]

Description

The String#crypt documentation states "the format and the result
are system and implementation dependent", so the tests should not
be checking for specific results. We should only be checking that
the method returns a string, that the results are different
if you use a different password or different salt, and the results
are the same if the using the encypted password as the salt
results in the same encrypted password.


Files

0001-Fix-tests-for-String-crypt.patch (2.02 KB) 0001-Fix-tests-for-String-crypt.patch jeremyevans0 (Jeremy Evans), 07/17/2015 09:29 PM

Updated by mame (Yusuke Endoh) almost 5 years ago

Do the tests cause any actual issue? Failed on OpenBSD?

Ruby's tests are not conformance test. They should include a test of implementation-defined behavior and it is actually useful to detect an unintentional change of behavior.

--
Yusuke Endoh mame@ruby-lang.org

Updated by jeremyevans0 (Jeremy Evans) almost 5 years ago

Well, the tests fail on OpenBSD regardless of the patch, as OpenBSD crypt(3) does not support the historical, insecure DES-based crypt, only bcrypt, and you need to pass a bcrypt salt as the second argument. I will be adding a local patch to OpenBSD so that passing a non-bcrypt salt to String#crypt will result in a bcrypt salt being autogenerated, so that portable ruby programs using String#crypt will run on OpenBSD. I don't think that patch belongs in ruby.

I think this patch improves things as it allows implementations to choose how to implement crypt(3), as long as they their crypt has the property:

enc = "pass".crypt("salt")
enc == "pass".crypt(enc)

This isn't a big deal, I'm fine keeping this as a local OpenBSD patch, but I think it makes sense to have this in ruby. If you disagree, feel free to close this issue.

Updated by jeremyevans0 (Jeremy Evans) about 1 year ago

  • Subject changed from Fix tests for String#crypt to Fix tests for String#crypt so they pass on OpenBSD

I've added a GitHub pull request to fix the String#crypt tests on OpenBSD (https://github.com/ruby/ruby/pull/2200). It's less invasive than the previous patch, and should result in the same behavior on !OpenBSD. I plan on merging it after I can do some testing on Linux, assuming there are no objections.

Updated by jeremyevans0 (Jeremy Evans) about 1 year ago

CI looks clean with https://github.com/ruby/ruby/pull/2200, so I'll be committing it in about a week if there are no objections.

#5

Updated by jeremyevans (Jeremy Evans) about 1 year ago

  • Status changed from Open to Closed

Applied in changeset git|4b9869e7e04cbfb581f6f1b21ae17b310135c5e2.


Update String#crypt tests to work on OpenBSD

Skip the webrick httpauth tests that use crypt when testing on
OpenBSD.

Fixes [Bug #11363]

Also available in: Atom PDF