On the redmine of my company, there is this bug where I get an internal error if I want to search into a project.
Here is the log corresponding to the error:
Processing by SearchController#index as HTML
Parameters: {"utf8"=>"✓", "issues"=>"1", "q"=>"test", "id"=>"sprint"}
Current user: me (id=60)
Completed 500 Internal Server Error in 85.0ms
TypeError (no implicit conversion of nil into String):
lib/plugins/acts_as_searchable/lib/acts_as_searchable.rb:126:in `search'
app/controllers/search_controller.rb:74:in `block in index'
app/controllers/search_controller.rb:73:in `each'
app/controllers/search_controller.rb:73:in `index'
The lines corresponding to the error in the controller are :
if !#tokens.empty?
# no more than 5 tokens to search for
#tokens.slice! 5..-1 if #tokens.size > 5
#results = []
#results_by_type = Hash.new {|h,k| h[k] = 0}
limit = 10
#scope.each do |s|
r, c = s.singularize.camelcase.constantize.search(#tokens, projects_to_search,
:all_words => #all_words,
:titles_only => #titles_only,
:limit => (limit+1),
:offset => offset,
:before => params[:previous].nil?)
#results += r
Here is my config :
Environment:
Redmine version 2.6.9.stable
Ruby version 2.3.0-p0 (2015-12-25) [x86_64-linux]
Rails version 3.2.22
Environment production
Database adapter PostgreSQL
SCM:
Git 1.9.1
Filesystem
Redmine plugins:
no plugin installed
What is interesting is that when I search only one letter, i'm redirected on the search page, but I don't have an internal error.
I'm very new to Redmine developpement and to Ruby, I was just assigned to try to fix this bug. Do any of you have an idea of how to fix it ?
Thanks.
I had the same issue, I was able to fix it by downgrading my ruby & rails version, a working set is :
Rails 3.2.19
Ruby 2.1.4p265
It's look like it's due to a braking change in ruby-2.3.0.
It's really odd. We are using 'Redmine' -2.0.3.1 and we do not have that kind of behaviour.
It seems that "someone" messed up the form linked to the search input. Try to follow that data from when you press enter until it answer you with the 500 error code.
If you can , change the environment to development, this way it will show you more detailed errors.
But I can't help you much more , you didn't provide enough info about the problem.
Related
I have been trying to upgrade my app from Rails 4 to Rails 5. In my Rails 4 version I have quiet_assets_path set but in Rails 5 it is not required. But when I removed that tried to start the server I am getting the following error,
> ruby-2.2.2/gems/rack-mini-profiler-0.10.2/lib/mini_profiler_rails/railtie.rb:93:in
> `>': comparison of Fixnum with nil failed (ArgumentError) from
> /Users/Admin/.rvm/gems/ruby-2.2.2/gems/rack-mini-profiler-0.10.2/lib/mini_profiler_rails/railtie.rb:93:in
> `block in <class:Railtie>'
Can someone help me with this?
Edit:
Following is my rack_profiler.rb,
if Rails.env.development? || Rails.env.production?
require 'rack-mini-profiler'
# initialization is skipped so trigger it
Rack::MiniProfilerRails.initialize!(Rails.application)
Rack::MiniProfiler.config.skip_schema_queries = true
Rack::MiniProfiler.config.skip_paths += %w(/admin/sidekiq)
Rails.application.middleware.delete(Rack::MiniProfiler)
Rails.application.middleware.insert_after(Rack::Deflater, Rack::MiniProfiler)
end
When I comment the delete line then server is starting but if the line uncommented then the server breaks.
thanks for the update. First of all, do you use Rack::Deflater middleware in development environment too?
I think this issue might help you. It basically says that in Rails all delete middleware operations are issued at the end. You can use the swap method as described in the above issue.
If you search the repo issues for "Deflater" you'll find a lot of results, but I believe the above contains your fix.
I have a Ruby 1.9.3 / Rails 3.1 project with the following in the gemfile:
gem 'rails', '3.1.12'
gem 'json'
gem 'multi_json', '1.7.7'
That version of rails sets activesupport to 3.1.12 as well. I'm not sure what the exact cause of the problem is, but when running bundle exec rake test, I got the error:
/home/user/.gem/ruby/1.9.3/gems/multi_json-1.7.7/lib/multi_json.rb:121:in 'rescue in load_adapter': Did not recognize your adapter specification. (ArgumentError)
...
(more stack trace, including activesupport methods)
Fortunately I found a solution! See below.
Edit: My original answer is outdated and incorrect; read it if you please, but please read the updated information at the bottom.
After viewing a ton of other questions such as these ones:
OmniAuth Login With Twitter - "Did not recognize your adapter specification." Error
Capistrano deploy: "Did not recognize your adapter specification" during assets:precompile
https://github.com/intridea/multi_json/issues/132
I hadn't found a solution, so I dove into the library and determined that load_adapter was receiving the parameter "JSONGem". The alias was failing, and the method attempted to load
/home/user/.gem/ruby/1.9.3/gems/multi_json-1.7.7/lib/multi_json/adapters/JSONGem.rb
This file doesn't exist, but .../json_gem.rb does exist! So I modified load adapter as follows:
def load_adapter(new_adapter)
# puts "new_adapter: #{new_adapter}" # Debugging
# puts "new_adapater.class: #{new_adapter.class}" # Debugging
case new_adapter
when String, Symbol
new_adapter = ALIASES.fetch(new_adapter.to_s, new_adapter)
new_adapter = "json_gem" if new_adapter =~ /^jsongem$/i # I added this line
# puts "final adapter: #{new_adapter}" # debugging
require "multi_json/adapters/#{new_adapter}"
klass_name = new_adapter.to_s.split('_').map(&:capitalize) * ''
MultiJson::Adapters.const_get(klass_name)
when NilClass, FalseClass
load_adapter default_adapter
when Class, Module
new_adapter
else
raise NameError
end
rescue NameError, ::LoadError
raise ArgumentError, 'Did not recognize your adapter specification.'
end
This fixed the problem for me. It's probably not an optimal solution (ideally I would understand WHY the ALIASES.fetch failed, if that is indeed what happened, and fix that), but if your problem is similar then hopefully this quick fix can help.
Update
It's not viable for deployability reasons to modify someone else's gem. Fortunately I found the root cause of the problem. In project_root/config/initializers/security_patches.rb, we had the line
ActiveSupport::JSON.backend = "JSONGem"
This was the recommended fix to a security bug in older versions of rails. Now that we are on a newer version of rails (i.e, > 3.0), we can simply replace "JSONGem" with "json_gem" (which is what my original modification was doing, in a roundabout way) and not worry about the security issue.
Is there an easy fix, how I could continue an old rails 3-0.20 installation under ruby 2.0?
The first error, caused by this line:
<%= stylesheet_link_tag :all %>
is
ActionView::Template::Error (no implicit conversion of nil into String):
An upgrade of the rails version would be the best, but unfortunately it is not possible in my case.
Hotfix the problem with the following line at the end in application.rb
ActionController::Base.config.relative_url_root = ''
I ran into the same issue. After drilling down into the stylesheet_link_tag method, I found that the issue comes from here
# actionpack-3.0.20/lib/action_view/helpers/asset_tag_helper.rb:749
if has_request && include_host && !source.start_with?(controller.config.relative_url_root)
The problem is String#starts_with?. In 1.9.3, that method will handle a nil as an input. 2.0.0 does not allow that.
ruby-1.9.3> 'whatever'.start_with? nil
=> false
ruby-2.0.0> 'whatever'.start_with? nil
TypeError: no implicit conversion of nil into String
It's probably also true that later versions of Rails set the value to '' if it's not set to prevent this issue. The hotfix mentioned above does fix the issue, but the root cause is differences between 1.9.3 and 2.0.0.
After upgrading to ruby 1.9 we began to notice pages failing to render from the rails template renderer when a user used a non-ASCII character. Specifically "é". I was able to resolve this issue on one of our staging servers, but I have not been able to reproduce the fix on our production server.
The fix that seemed to work the first time:
Converted the database from latin1 to utf8 using the convert_charset tool available here: http://www.mysqlperformanceblog.com/2009/03/17/converting-character-sets/. (including setting default_character_set=utf8 in my.cnf and running SET GLOBAL character_set_server=utf8
Switched to the sam-mysql-ruby adapter (instead of the standard mysql adapter: http://gemcutter.org/gems/sam-mysql-ruby)
Restarted rails
The error is:
"invalid byte sequence in US-ASCII"
Oddly, after following the steps above the error has not changed on our production server. Setting encoding: utf8 in database.yml does not change the error either.
The error raised on the following line of code:
<%= link_to h(question.title), question_path(question) %>
This blog seems to suggest a fix, but it mentions that this should not be a problem in 1.9: http://www.igvita.com/2007/04/11/secure-utf-8-input-in-rails/ (and it's over 2 years old).
I imagine this problem might soon affect a lot of people as more rails developers people switch to 1.9.
I found the solution:
The problem is:
Fetching data from any database (Mysql, Postgresql, Sqlite2 & 3), all configured to have UTF-8 as it's character set, returns the data with ASCII-8BIT in ruby 1.9.1 and rails 2.3.2.1.
(Taken from: https://rails.lighthouseapp.com/projects/8994/tickets/2476)
My attempt to use the patched mysql adapter likely failed because my database was not configured to natively use utf8, so the patched adapter failed to work properly.
The fix ended up being to use the patch file available here: http://gnuu.org/2009/11/06/ruby19-rails-mysql-utf8/
require 'mysql'
class Mysql::Result
def encode(value, encoding = "utf-8")
String === value ? value.force_encoding(encoding) : value
end
def each_utf8(&block)
each_orig do |row|
yield row.map {|col| encode(col) }
end
end
alias each_orig each
alias each each_utf8
def each_hash_utf8(&block)
each_hash_orig do |row|
row.each {|k, v| row[k] = encode(v) }
yield(row)
end
end
alias each_hash_orig each_hash
alias each_hash each_hash_utf8
end
(Placed in lib/mysql_utf8fix.rb and required in enviornment.rb using require 'lib/mysql_utf8fix.rb')
it is only require 'mysql_utf8fix.rb' (rails 2.3.11)
Please user mysql2(gem) adapter instead of mysql adapter in database.yml
and remove the mysql patches(If exists) and add the following lines in environment.rb.
Encoding.default_external = Encoding::UTF_8
Encoding.default_internal = Encoding::UTF_8
Then run in apache and passenger it ll work fine
Thanks,
Ramanavel Selvaraju.
Ok, so we're upgrading a client's legacy code from 2.0.2 to latest rails. Most of the basics were easy to fix, but I can't get to the admin screens. Every time we hit "current_user" we get a "stack level too deep" error.
I've dug deeply into the code (read: flailed around a lot at random) and I've finally narrowed it down to the ActiveRecord::Session store.
The code berks out on the line that includes "session[:user]".
When I spool up script/console I can replicate the stack overflow with the following line:
s = ActiveRecord::SessionStore::Session.new(:session_id => '42', :data => {})
stacktrace to follow.
To make sure it wasn't some weird incompatibility, I blew away the session table in the db and reloaded using rake db:sessions:create and it's still asploding on that line.
SystemStackError: stack level too deep
from /usr/lib/ruby/gems/1.8/gems/activerecord-2.3.5/lib/active_record/connection_adapters/abstract/schema_definitions.rb:68:in `type_cast'
from /usr/lib/ruby/gems/1.8/gems/activerecord-2.3.5/lib/active_record/dirty.rb:161:in `field_changed?'
from /usr/lib/ruby/gems/1.8/gems/activerecord-2.3.5/lib/active_record/dirty.rb:132:in `write_attribute_without_dirty'
from /usr/lib/ruby/gems/1.8/gems/activerecord-2.3.5/lib/active_record/dirty.rb:139:in `write_attribute'
from /usr/lib/ruby/gems/1.8/gems/activerecord-2.3.5/lib/active_record/attribute_methods.rb:211:in `session_id='
from /usr/lib/ruby/gems/1.8/gems/activerecord-2.3.5/lib/active_record/base.rb:2746:in `send'
from /usr/lib/ruby/gems/1.8/gems/activerecord-2.3.5/lib/active_record/base.rb:2746:in `attributes='
from /usr/lib/ruby/gems/1.8/gems/activerecord-2.3.5/lib/active_record/base.rb:2742:in `each'
from /usr/lib/ruby/gems/1.8/gems/activerecord-2.3.5/lib/active_record/base.rb:2742:in `attributes='
from /usr/lib/ruby/gems/1.8/gems/activerecord-2.3.5/lib/active_record/base.rb:2438:in `initialize'
from (irb):10:in `new'
I went into active_record and put in heaps of puts lines. Here's a brief version the final two lines are continually repeated and are clearly the stack-overflow... but why does column_for_attribute(session_id) cause an overflow?
"in respond_to? method: session_id="
"in respond_to? session_id= - there are no generated methods. generating..."
"in respond_to? session_id= - methods generated"
"responds to session_id=? y"
"in respond_to? method: session_id="
"in respond_to? session_id= - super is true"
"in read_attribute: session_id about to fetch value"
"fetched attr value: nil"
"read attr: session_id - value nil"
"column for attribute: session_id"
"c_for_a got col hash - accessing for name: session_id"
"column for attribute: session_id"
"c_for_a got col hash - accessing for name: session_id"
"column for attribute: session_id"
"c_for_a got col hash - accessing for name: session_id"
"column for attribute: session_id"
"c_for_a got col hash - accessing for name: session_id"
Note: we also tried upgrading just to 2.1.0 and the same error occurs (when trying to login - obviously the session-class changed names in between so the scipt/console issue depends on the classname). This time it's endlessly trying just "id" instead of "session_id".
Current environment.rb session settings (for the 2.1.0 version) is below:
# Use the database for sessions instead of the file system
# (create the session table with 'rake db:sessions:create')
config.action_controller.session_store = :active_record_store
config.action_controller.session = { :session_key => "_our_session_id",
:secret=> "some random secret key of your choosing over 30 characters" }
If I drop it back to v2.0.2 it works again.
So I guess the question is: what changed in the ActiveRecord Session between v 2.0.2 and 2.1.0 and how do I make it work?
Ok, looks like the problem was an old back-port/monkey-patch. In the plugins directory we had a back-port of the "dirty" functionality - which I'm guessing must've come out in Rails 2.1.0. All I know is that deleting that plugin seems to have made the "stack level too deep" problem go away (for now). :)