Commit 4345bb8c authored by Douwe Maan's avatar Douwe Maan

Fix ambiguous routing issues by teaching router about reserved words

parent 3cfcbcf3
...@@ -74,6 +74,6 @@ class Projects::RefsController < Projects::ApplicationController ...@@ -74,6 +74,6 @@ class Projects::RefsController < Projects::ApplicationController
private private
def validate_ref_id def validate_ref_id
return not_found! if params[:id].present? && params[:id] !~ Gitlab::Regex.git_reference_regex return not_found! if params[:id].present? && params[:id] !~ Gitlab::PathRegex.git_reference_regex
end end
end end
...@@ -205,8 +205,8 @@ class Project < ActiveRecord::Base ...@@ -205,8 +205,8 @@ class Project < ActiveRecord::Base
presence: true, presence: true,
dynamic_path: true, dynamic_path: true,
length: { maximum: 255 }, length: { maximum: 255 },
format: { with: Gitlab::Regex.project_path_regex, format: { with: Gitlab::PathRegex.project_path_format_regex,
message: Gitlab::Regex.project_path_regex_message }, message: Gitlab::PathRegex.project_path_format_message },
uniqueness: { scope: :namespace_id } uniqueness: { scope: :namespace_id }
validates :namespace, presence: true validates :namespace, presence: true
...@@ -380,11 +380,9 @@ class Project < ActiveRecord::Base ...@@ -380,11 +380,9 @@ class Project < ActiveRecord::Base
end end
def reference_pattern def reference_pattern
name_pattern = Gitlab::Regex::FULL_NAMESPACE_REGEX_STR
%r{ %r{
((?<namespace>#{name_pattern})\/)? ((?<namespace>#{Gitlab::PathRegex::FULL_NAMESPACE_FORMAT_REGEX})\/)?
(?<project>#{name_pattern}) (?<project>#{Gitlab::PathRegex::PROJECT_PATH_FORMAT_REGEX})
}x }x
end end
......
...@@ -367,7 +367,7 @@ class User < ActiveRecord::Base ...@@ -367,7 +367,7 @@ class User < ActiveRecord::Base
def reference_pattern def reference_pattern
%r{ %r{
#{Regexp.escape(reference_prefix)} #{Regexp.escape(reference_prefix)}
(?<user>#{Gitlab::Regex::FULL_NAMESPACE_REGEX_STR}) (?<user>#{Gitlab::PathRegex::FULL_NAMESPACE_FORMAT_REGEX})
}x }x
end end
......
...@@ -3,212 +3,45 @@ ...@@ -3,212 +3,45 @@
# Custom validator for GitLab path values. # Custom validator for GitLab path values.
# These paths are assigned to `Namespace` (& `Group` as a subclass) & `Project` # These paths are assigned to `Namespace` (& `Group` as a subclass) & `Project`
# #
# Values are checked for formatting and exclusion from a list of reserved path # Values are checked for formatting and exclusion from a list of illegal path
# names. # names.
class DynamicPathValidator < ActiveModel::EachValidator class DynamicPathValidator < ActiveModel::EachValidator
# All routes that appear on the top level must be listed here. class << self
# This will make sure that groups cannot be created with these names def valid_user_path?(path)
# as these routes would be masked by the paths already in place. "#{path}/" =~ Gitlab::PathRegex.root_namespace_path_regex
# end
# Example:
# /api/api-project
#
# the path `api` shouldn't be allowed because it would be masked by `api/*`
#
TOP_LEVEL_ROUTES = %w[
-
.well-known
abuse_reports
admin
all
api
assets
autocomplete
ci
dashboard
explore
files
groups
health_check
help
hooks
import
invites
issues
jwt
koding
member
merge_requests
new
notes
notification_settings
oauth
profile
projects
public
repository
robots.txt
s
search
sent_notifications
services
snippets
teams
u
unicorn_test
unsubscribes
uploads
users
].freeze
# This list should contain all words following `/*namespace_id/:project_id` in
# routes that contain a second wildcard.
#
# Example:
# /*namespace_id/:project_id/badges/*ref/build
#
# If `badges` was allowed as a project/group name, we would not be able to access the
# `badges` route for those projects:
#
# Consider a namespace with path `foo/bar` and a project called `badges`.
# The route to the build badge would then be `/foo/bar/badges/badges/master/build.svg`
#
# When accessing this path the route would be matched to the `badges` path
# with the following params:
# - namespace_id: `foo`
# - project_id: `bar`
# - ref: `badges/master`
#
# Failing to find the project, this would result in a 404.
#
# By rejecting `badges` the router can _count_ on the fact that `badges` will
# be preceded by the `namespace/project`.
WILDCARD_ROUTES = %w[
badges
blame
blob
builds
commits
create
create_dir
edit
environments/folders
files
find_file
gitlab-lfs/objects
info/lfs/objects
new
preview
raw
refs
tree
update
wikis
].freeze
# These are all the paths that follow `/groups/*id/ or `/groups/*group_id`
# We need to reject these because we have a `/groups/*id` page that is the same
# as the `/*id`.
#
# If we would allow a subgroup to be created with the name `activity` then
# this group would not be accessible through `/groups/parent/activity` since
# this would map to the activity-page of it's parent.
GROUP_ROUTES = %w[
activity
analytics
audit_events
avatar
edit
group_members
hooks
issues
labels
ldap
ldap_group_links
merge_requests
milestones
notification_setting
pipeline_quota
projects
subgroups
].freeze
CHILD_ROUTES = (WILDCARD_ROUTES | GROUP_ROUTES).freeze
def self.without_reserved_wildcard_paths_regex
@without_reserved_wildcard_paths_regex ||= regex_excluding_child_paths(WILDCARD_ROUTES)
end
def self.without_reserved_child_paths_regex
@without_reserved_child_paths_regex ||= regex_excluding_child_paths(CHILD_ROUTES)
end
# This is used to validate a full path.
# It doesn't match paths
# - Starting with one of the top level words
# - Containing one of the child level words in the middle of a path
def self.regex_excluding_child_paths(child_routes)
reserved_top_level_words = Regexp.union(TOP_LEVEL_ROUTES)
not_starting_in_reserved_word = %r{\A/?(?!(#{reserved_top_level_words})(/|\z))}
reserved_child_level_words = Regexp.union(child_routes)
not_containing_reserved_child = %r{(?!\S+/(#{reserved_child_level_words})(/|\z))}
%r{#{not_starting_in_reserved_word}
#{not_containing_reserved_child}
#{Gitlab::Regex.full_namespace_regex}}x
end
def self.valid?(path)
path =~ Gitlab::Regex.full_namespace_regex && !full_path_reserved?(path)
end
def self.full_path_reserved?(path)
path = path.to_s.downcase
_project_part, namespace_parts = path.reverse.split('/', 2).map(&:reverse)
wildcard_reserved?(path) || child_reserved?(namespace_parts)
end
def self.child_reserved?(path)
return false unless path
path !~ without_reserved_child_paths_regex
end
def self.wildcard_reserved?(path) def valid_group_path?(path)
return false unless path "#{path}/" =~ Gitlab::PathRegex.full_namespace_path_regex
end
path !~ without_reserved_wildcard_paths_regex def valid_project_path?(path)
"#{path}/" =~ Gitlab::PathRegex.full_project_path_regex
end
end end
delegate :full_path_reserved?, def path_valid_for_record?(record, value)
:child_reserved?,
to: :class
def path_reserved_for_record?(record, value)
full_path = record.respond_to?(:full_path) ? record.full_path : value full_path = record.respond_to?(:full_path) ? record.full_path : value
# For group paths the entire path cannot contain a reserved child word return true unless full_path
# The path doesn't contain the last `_project_part` so we need to validate
# if the entire path. case record
# Example: when Project
# A *group* with full path `parent/activity` is reserved. self.class.valid_project_path?(full_path)
# A *project* with full path `parent/activity` is allowed. when Group
if record.is_a? Group self.class.valid_group_path?(full_path)
child_reserved?(full_path) else # User or non-Group Namespace
else self.class.valid_user_path?(full_path)
full_path_reserved?(full_path)
end end
end end
def validate_each(record, attribute, value) def validate_each(record, attribute, value)
unless value =~ Gitlab::Regex.namespace_regex unless value =~ Gitlab::PathRegex.namespace_format_regex
record.errors.add(attribute, Gitlab::Regex.namespace_regex_message) record.errors.add(attribute, Gitlab::PathRegex.namespace_format_message)
return return
end end
if path_reserved_for_record?(record, value) unless path_valid_for_record?(record, value)
record.errors.add(attribute, "#{value} is a reserved name") record.errors.add(attribute, "#{value} is a reserved name")
end end
end end
......
...@@ -8,7 +8,7 @@ ...@@ -8,7 +8,7 @@
= f.text_field :name, class: "form-control top", required: true, title: "This field is required." = f.text_field :name, class: "form-control top", required: true, title: "This field is required."
.username.form-group .username.form-group
= f.label :username = f.label :username
= f.text_field :username, class: "form-control middle", pattern: Gitlab::Regex::NAMESPACE_REGEX_STR_JS, required: true, title: 'Please create a username with only alphanumeric characters.' = f.text_field :username, class: "form-control middle", pattern: Gitlab::PathRegex::NAMESPACE_FORMAT_REGEX_JS, required: true, title: 'Please create a username with only alphanumeric characters.'
%p.validation-error.hide Username is already taken. %p.validation-error.hide Username is already taken.
%p.validation-success.hide Username is available. %p.validation-success.hide Username is available.
%p.validation-pending.hide Checking username availability... %p.validation-pending.hide Checking username availability...
......
...@@ -15,7 +15,7 @@ ...@@ -15,7 +15,7 @@
%strong= parent.full_path + '/' %strong= parent.full_path + '/'
= f.text_field :path, placeholder: 'open-source', class: 'form-control', = f.text_field :path, placeholder: 'open-source', class: 'form-control',
autofocus: local_assigns[:autofocus] || false, required: true, autofocus: local_assigns[:autofocus] || false, required: true,
pattern: Gitlab::Regex::NAMESPACE_REGEX_STR_JS, pattern: Gitlab::PathRegex::NAMESPACE_FORMAT_REGEX_JS,
title: 'Please choose a group path with no special characters.', title: 'Please choose a group path with no special characters.',
"data-bind-in" => "#{'create_chat_team' if Gitlab.config.mattermost.enabled}" "data-bind-in" => "#{'create_chat_team' if Gitlab.config.mattermost.enabled}"
- if parent - if parent
......
require 'sidekiq/web' require 'sidekiq/web'
require 'sidekiq/cron/web' require 'sidekiq/cron/web'
require 'constraints/group_url_constrainer'
Rails.application.routes.draw do Rails.application.routes.draw do
concern :access_requestable do concern :access_requestable do
...@@ -85,20 +84,6 @@ Rails.application.routes.draw do ...@@ -85,20 +84,6 @@ Rails.application.routes.draw do
root to: "root#index" root to: "root#index"
# Since group show page is wildcard routing
# we want all other routing to be checked before matching this one
constraints(GroupUrlConstrainer.new) do
scope(path: '*id',
as: :group,
constraints: { id: Gitlab::Regex.namespace_route_regex, format: /(html|json|atom)/ },
controller: :groups) do
get '/', action: :show
patch '/', action: :update
put '/', action: :update
delete '/', action: :destroy
end
end
draw :test if Rails.env.test? draw :test if Rails.env.test?
get '*unmatched_route', to: 'application#route_not_found' get '*unmatched_route', to: 'application#route_not_found'
......
...@@ -36,7 +36,7 @@ namespace :admin do ...@@ -36,7 +36,7 @@ namespace :admin do
scope(path: 'groups/*id', scope(path: 'groups/*id',
controller: :groups, controller: :groups,
constraints: { id: Gitlab::Regex.namespace_route_regex, format: /(html|json|atom)/ }) do constraints: { id: Gitlab::PathRegex.full_namespace_route_regex, format: /(html|json|atom)/ }) do
scope(as: :group) do scope(as: :group) do
put :members_update put :members_update
...@@ -68,10 +68,12 @@ namespace :admin do ...@@ -68,10 +68,12 @@ namespace :admin do
resources :projects, only: [:index] resources :projects, only: [:index]
scope(path: 'projects/*namespace_id', as: :namespace) do scope(path: 'projects/*namespace_id',
as: :namespace,
constraints: { namespace_id: Gitlab::PathRegex.full_namespace_route_regex }) do
resources(:projects, resources(:projects,
path: '/', path: '/',
constraints: { id: Gitlab::Regex.project_route_regex }, constraints: { id: Gitlab::PathRegex.project_route_regex },
only: [:show]) do only: [:show]) do
member do member do
......
scope(path: '*namespace_id/:project_id', constraints: { format: nil }) do scope(path: '*namespace_id/:project_id',
scope(constraints: { project_id: Gitlab::Regex.project_git_route_regex }, module: :projects) do format: nil,
constraints: { namespace_id: Gitlab::PathRegex.full_namespace_route_regex }) do
scope(constraints: { project_id: Gitlab::PathRegex.project_git_route_regex }, module: :projects) do
# Git HTTP clients ('git clone' etc.) # Git HTTP clients ('git clone' etc.)
scope(controller: :git_http) do scope(controller: :git_http) do
get '/info/refs', action: :info_refs get '/info/refs', action: :info_refs
...@@ -26,7 +28,7 @@ scope(path: '*namespace_id/:project_id', constraints: { format: nil }) do ...@@ -26,7 +28,7 @@ scope(path: '*namespace_id/:project_id', constraints: { format: nil }) do
end end
# Redirect /group/project/info/refs to /group/project.git/info/refs # Redirect /group/project/info/refs to /group/project.git/info/refs
scope(constraints: { project_id: Gitlab::Regex.project_route_regex }) do scope(constraints: { project_id: Gitlab::PathRegex.project_route_regex }) do
# Allow /info/refs, /info/refs?service=git-upload-pack, and # Allow /info/refs, /info/refs?service=git-upload-pack, and
# /info/refs?service=git-receive-pack, but nothing else. # /info/refs?service=git-receive-pack, but nothing else.
# #
......
require 'constraints/group_url_constrainer'
resources :groups, only: [:index, :new, :create] resources :groups, only: [:index, :new, :create]
scope(path: 'groups/*group_id', scope(path: 'groups/*group_id',
module: :groups, module: :groups,
as: :group, as: :group,
constraints: { group_id: Gitlab::Regex.namespace_route_regex }) do constraints: { group_id: Gitlab::PathRegex.full_namespace_route_regex }) do
resources :group_members, only: [:index, :create, :update, :destroy], concerns: :access_requestable do resources :group_members, only: [:index, :create, :update, :destroy], concerns: :access_requestable do
post :resend_invite, on: :member post :resend_invite, on: :member
delete :leave, on: :collection delete :leave, on: :collection
...@@ -25,7 +27,7 @@ end ...@@ -25,7 +27,7 @@ end
scope(path: 'groups/*id', scope(path: 'groups/*id',
controller: :groups, controller: :groups,
constraints: { id: Gitlab::Regex.namespace_route_regex, format: /(html|json|atom)/ }) do constraints: { id: Gitlab::PathRegex.full_namespace_route_regex, format: /(html|json|atom)/ }) do
get :edit, as: :edit_group get :edit, as: :edit_group
get :issues, as: :issues_group get :issues, as: :issues_group
get :merge_requests, as: :merge_requests_group get :merge_requests, as: :merge_requests_group
...@@ -34,3 +36,15 @@ scope(path: 'groups/*id', ...@@ -34,3 +36,15 @@ scope(path: 'groups/*id',
get :subgroups, as: :subgroups_group get :subgroups, as: :subgroups_group
get '/', action: :show, as: :group_canonical get '/', action: :show, as: :group_canonical
end end
constraints(GroupUrlConstrainer.new) do
scope(path: '*id',
as: :group,
constraints: { id: Gitlab::PathRegex.full_namespace_route_regex, format: /(html|json|atom)/ },
controller: :groups) do
get '/', action: :show
patch '/', action: :update
put '/', action: :update
delete '/', action: :destroy
end
end
...@@ -5,9 +5,24 @@ resources :projects, only: [:index, :new, :create] ...@@ -5,9 +5,24 @@ resources :projects, only: [:index, :new, :create]
draw :git_http draw :git_http
constraints(ProjectUrlConstrainer.new) do constraints(ProjectUrlConstrainer.new) do
scope(path: '*namespace_id', as: :namespace) do # If the route has a wildcard segment, the segment has a regex constraint,
# the segment is potentially followed by _another_ wildcard segment, and
# the `format` option is not set to false, we need to specify that
# regex constraint _outside_ of `constraints: {}`.
#
# Otherwise, Rails will overwrite the constraint with `/.+?/`,
# which breaks some of our wildcard routes like `/blob/*id`
# and `/tree/*id` that depend on the negative lookahead inside
# `Gitlab::PathRegex.full_namespace_route_regex`, which helps the router
# determine whether a certain path segment is part of `*namespace_id`,
# `:project_id`, or `*id`.
#
# See https://github.com/rails/rails/blob/v4.2.8/actionpack/lib/action_dispatch/routing/mapper.rb#L155
scope(path: '*namespace_id',
as: :namespace,
namespace_id: Gitlab::PathRegex.full_namespace_route_regex) do
scope(path: ':project_id', scope(path: ':project_id',
constraints: { project_id: Gitlab::Regex.project_route_regex }, constraints: { project_id: Gitlab::PathRegex.project_route_regex },
module: :projects, module: :projects,
as: :project) do as: :project) do
...@@ -314,7 +329,7 @@ constraints(ProjectUrlConstrainer.new) do ...@@ -314,7 +329,7 @@ constraints(ProjectUrlConstrainer.new) do
resources :runner_projects, only: [:create, :destroy] resources :runner_projects, only: [:create, :destroy]
resources :badges, only: [:index] do resources :badges, only: [:index] do
collection do collection do
scope '*ref', constraints: { ref: Gitlab::Regex.git_reference_regex } do scope '*ref', constraints: { ref: Gitlab::PathRegex.git_reference_regex } do
constraints format: /svg/ do constraints format: /svg/ do
get :build get :build
get :coverage get :coverage
...@@ -337,7 +352,7 @@ constraints(ProjectUrlConstrainer.new) do ...@@ -337,7 +352,7 @@ constraints(ProjectUrlConstrainer.new) do
resources(:projects, resources(:projects,
path: '/', path: '/',
constraints: { id: Gitlab::Regex.project_route_regex }, constraints: { id: Gitlab::PathRegex.project_route_regex },
only: [:edit, :show, :update, :destroy]) do only: [:edit, :show, :update, :destroy]) do
member do member do
put :transfer put :transfer
......
...@@ -2,7 +2,7 @@ ...@@ -2,7 +2,7 @@
resource :repository, only: [:create] do resource :repository, only: [:create] do
member do member do
get 'archive', constraints: { format: Gitlab::Regex.archive_formats_regex } get 'archive', constraints: { format: Gitlab::PathRegex.archive_formats_regex }
end end
end end
...@@ -24,7 +24,7 @@ scope format: false do ...@@ -24,7 +24,7 @@ scope format: false do
member do member do
# tree viewer logs # tree viewer logs
get 'logs_tree', constraints: { id: Gitlab::Regex.git_reference_regex } get 'logs_tree', constraints: { id: Gitlab::PathRegex.git_reference_regex }
# Directories with leading dots erroneously get rejected if git # Directories with leading dots erroneously get rejected if git
# ref regex used in constraints. Regex verification now done in controller. # ref regex used in constraints. Regex verification now done in controller.
get 'logs_tree/*path', action: :logs_tree, as: :logs_file, format: false, constraints: { get 'logs_tree/*path', action: :logs_tree, as: :logs_file, format: false, constraints: {
...@@ -34,7 +34,7 @@ scope format: false do ...@@ -34,7 +34,7 @@ scope format: false do
end end
end end
scope constraints: { id: Gitlab::Regex.git_reference_regex } do scope constraints: { id: Gitlab::PathRegex.git_reference_regex } do
resources :network, only: [:show] resources :network, only: [:show]
resources :graphs, only: [:show] do resources :graphs, only: [:show] do
......
...@@ -11,19 +11,7 @@ devise_scope :user do ...@@ -11,19 +11,7 @@ devise_scope :user do
get '/users/almost_there' => 'confirmations#almost_there' get '/users/almost_there' => 'confirmations#almost_there'
end end
constraints(UserUrlConstrainer.new) do scope(constraints: { username: Gitlab::PathRegex.root_namespace_route_regex }) do
# Get all keys of user
get ':username.keys' => 'profiles/keys#get_keys', constraints: { username: Gitlab::Regex.namespace_route_regex }
scope(path: ':username',
as: :user,
constraints: { username: Gitlab::Regex.namespace_route_regex },
controller: :users) do
get '/', action: :show
end
end
scope(constraints: { username: Gitlab::Regex.namespace_route_regex }) do
scope(path: 'users/:username', scope(path: 'users/:username',
as: :user, as: :user,
controller: :users) do controller: :users) do
...@@ -46,3 +34,15 @@ scope(constraints: { username: Gitlab::Regex.namespace_route_regex }) do ...@@ -46,3 +34,15 @@ scope(constraints: { username: Gitlab::Regex.namespace_route_regex }) do
get '/u/:username/snippets', to: redirect('/users/%{username}/snippets') get '/u/:username/snippets', to: redirect('/users/%{username}/snippets')
get '/u/:username/contributed', to: redirect('/users/%{username}/contributed') get '/u/:username/contributed', to: redirect('/users/%{username}/contributed')
end end
constraints(UserUrlConstrainer.new) do
# Get all keys of user
get ':username.keys' => 'profiles/keys#get_keys', constraints: { username: Gitlab::PathRegex.root_namespace_route_regex }
scope(path: ':username',
as: :user,
constraints: { username: Gitlab::PathRegex.root_namespace_route_regex },
controller: :users) do
get '/', action: :show
end
end
...@@ -71,9 +71,9 @@ structure. ...@@ -71,9 +71,9 @@ structure.
- You need to be an Owner of a group in order to be able to create - You need to be an Owner of a group in order to be able to create
a subgroup. For more information check the [permissions table][permissions]. a subgroup. For more information check the [permissions table][permissions].
- For a list of words that are not allowed to be used as group names see the - For a list of words that are not allowed to be used as group names see the
[`dynamic_path_validator.rb` file][reserved] under the `TOP_LEVEL_ROUTES`, `WILDCARD_ROUTES` and `GROUP_ROUTES` lists: [`dynamic_path_validator.rb` file][reserved] under the `TOP_LEVEL_ROUTES`, `PROJECT_WILDCARD_ROUTES` and `GROUP_ROUTES` lists:
- `TOP_LEVEL_ROUTES`: are names that are reserved as usernames or top level groups - `TOP_LEVEL_ROUTES`: are names that are reserved as usernames or top level groups
- `WILDCARD_ROUTES`: are names that are reserved for child groups or projects. - `PROJECT_WILDCARD_ROUTES`: are names that are reserved for child groups or projects.
- `GROUP_ROUTES`: are names that are reserved for all groups or projects. - `GROUP_ROUTES`: are names that are reserved for all groups or projects.
To create a subgroup: To create a subgroup:
......
...@@ -85,7 +85,7 @@ module API ...@@ -85,7 +85,7 @@ module API
optional :sha, type: String, desc: 'The commit sha of the archive to be downloaded' optional :sha, type: String, desc: 'The commit sha of the archive to be downloaded'
optional :format, type: String, desc: 'The archive format' optional :format, type: String, desc: 'The archive format'
end end
get ':id/repository/archive', requirements: { format: Gitlab::Regex.archive_formats_regex } do get ':id/repository/archive', requirements: { format: Gitlab::PathRegex.archive_formats_regex } do
begin begin
send_git_archive user_project.repository, ref: params[:sha], format: params[:format] send_git_archive user_project.repository, ref: params[:sha], format: params[:format]
rescue rescue
......
...@@ -72,7 +72,7 @@ module API ...@@ -72,7 +72,7 @@ module API
optional :sha, type: String, desc: 'The commit sha of the archive to be downloaded' optional :sha, type: String, desc: 'The commit sha of the archive to be downloaded'
optional :format, type: String, desc: 'The archive format' optional :format, type: String, desc: 'The archive format'
end end
get ':id/repository/archive', requirements: { format: Gitlab::Regex.archive_formats_regex } do get ':id/repository/archive', requirements: { format: Gitlab::PathRegex.archive_formats_regex } do
begin begin
send_git_archive user_project.repository, ref: params[:sha], format: params[:format] send_git_archive user_project.repository, ref: params[:sha], format: params[:format]
rescue rescue
......
class GroupUrlConstrainer class GroupUrlConstrainer
def matches?(request) def matches?(request)
id = request.params[:id] full_path = request.params[:group_id] || request.params[:id]
return false unless DynamicPathValidator.valid?(id) return false unless DynamicPathValidator.valid_group_path?(full_path)
Group.find_by_full_path(id, follow_redirects: request.get?).present? Group.find_by_full_path(full_path, follow_redirects: request.get?).present?
end end
end end
...@@ -2,9 +2,9 @@ class ProjectUrlConstrainer ...@@ -2,9 +2,9 @@ class ProjectUrlConstrainer
def matches?(request) def matches?(request)
namespace_path = request.params[:namespace_id] namespace_path = request.params[:namespace_id]
project_path = request.params[:project_id] || request.params[:id] project_path = request.params[:project_id] || request.params[:id]
full_path = namespace_path + '/' + project_path full_path = [namespace_path, project_path].join('/')
return false unless DynamicPathValidator.valid?(full_path) return false unless DynamicPathValidator.valid_project_path?(full_path)
Project.find_by_full_path(full_path, follow_redirects: request.get?).present? Project.find_by_full_path(full_path, follow_redirects: request.get?).present?
end end
......
class UserUrlConstrainer class UserUrlConstrainer
def matches?(request) def matches?(request)
User.find_by_full_path(request.params[:username], follow_redirects: request.get?).present? full_path = request.params[:username]
return false unless DynamicPathValidator.valid_user_path?(full_path)
User.find_by_full_path(full_path, follow_redirects: request.get?).present?
end end
end end
...@@ -10,7 +10,7 @@ module Gitlab ...@@ -10,7 +10,7 @@ module Gitlab
# - Ending in `issues/id`/realtime_changes` for the `issue_title` route # - Ending in `issues/id`/realtime_changes` for the `issue_title` route
USED_IN_ROUTES = %w[noteable issue notes issues realtime_changes USED_IN_ROUTES = %w[noteable issue notes issues realtime_changes
commit pipelines merge_requests new].freeze commit pipelines merge_requests new].freeze
RESERVED_WORDS = DynamicPathValidator::WILDCARD_ROUTES - USED_IN_ROUTES RESERVED_WORDS = Gitlab::PathRegex::ILLEGAL_PROJECT_PATH_WORDS - USED_IN_ROUTES
RESERVED_WORDS_REGEX = Regexp.union(*RESERVED_WORDS) RESERVED_WORDS_REGEX = Regexp.union(*RESERVED_WORDS)
ROUTES = [ ROUTES = [
Gitlab::EtagCaching::Router::Route.new( Gitlab::EtagCaching::Router::Route.new(
......
module Gitlab
module PathRegex
extend self
# All routes that appear on the top level must be listed here.
# This will make sure that groups cannot be created with these names
# as these routes would be masked by the paths already in place.
#
# Example:
# /api/api-project
#
# the path `api` shouldn't be allowed because it would be masked by `api/*`
#
TOP_LEVEL_ROUTES = %w[
-
.well-known
abuse_reports
admin
all
api
assets
autocomplete
ci
dashboard
explore
files
groups
health_check
help
hooks
import
invites
issues
jwt
koding
member
merge_requests
new
notes
notification_settings
oauth
profile
projects
public
repository
robots.txt
s
search
sent_notifications
services
snippets
teams
u
unicorn_test
unsubscribes
uploads
users
].freeze
# This list should contain all words following `/*namespace_id/:project_id` in
# routes that contain a second wildcard.
#
# Example:
# /*namespace_id/:project_id/badges/*ref/build
#
# If `badges` was allowed as a project/group name, we would not be able to access the
# `badges` route for those projects:
#
# Consider a namespace with path `foo/bar` and a project called `badges`.
# The route to the build badge would then be `/foo/bar/badges/badges/master/build.svg`
#
# When accessing this path the route would be matched to the `badges` path
# with the following params:
# - namespace_id: `foo`
# - project_id: `bar`
# - ref: `badges/master`
#
# Failing to find the project, this would result in a 404.
#
# By rejecting `badges` the router can _count_ on the fact that `badges` will
# be preceded by the `namespace/project`.
PROJECT_WILDCARD_ROUTES = %w[
badges
blame
blob
builds
commits
create
create_dir
edit
environments/folders
files
find_file
gitlab-lfs/objects
info/lfs/objects
new
preview
raw
refs
tree
update
wikis
].freeze
# These are all the paths that follow `/groups/*id/ or `/groups/*group_id`
# We need to reject these because we have a `/groups/*id` page that is the same
# as the `/*id`.
#
# If we would allow a subgroup to be created with the name `activity` then
# this group would not be accessible through `/groups/parent/activity` since
# this would map to the activity-page of its parent.
GROUP_ROUTES = %w[
activity
analytics
audit_events
avatar
edit
group_members
hooks
issues
labels
ldap
ldap_group_links
merge_requests
milestones
notification_setting
pipeline_quota
projects
subgroups
].freeze
ILLEGAL_PROJECT_PATH_WORDS = PROJECT_WILDCARD_ROUTES
ILLEGAL_GROUP_PATH_WORDS = (PROJECT_WILDCARD_ROUTES | GROUP_ROUTES).freeze
# The namespace regex is used in JavaScript to validate usernames in the "Register" form. However, Javascript
# does not support the negative lookbehind assertion (?<!) that disallows usernames ending in `.git` and `.atom`.
# Since this is a non-trivial problem to solve in Javascript (heavily complicate the regex, modify view code to
# allow non-regex validations, etc), `NAMESPACE_FORMAT_REGEX_JS` serves as a Javascript-compatible version of
# `NAMESPACE_FORMAT_REGEX`, with the negative lookbehind assertion removed. This means that the client-side validation
# will pass for usernames ending in `.atom` and `.git`, but will be caught by the server-side validation.
PATH_REGEX_STR = '[a-zA-Z0-9_\.][a-zA-Z0-9_\-\.]*'.freeze
NAMESPACE_FORMAT_REGEX_JS = PATH_REGEX_STR + '[a-zA-Z0-9_\-]|[a-zA-Z0-9_]'.freeze
NO_SUFFIX_REGEX = /(?<!\.git|\.atom)/.freeze
NAMESPACE_FORMAT_REGEX = /(?:#{NAMESPACE_FORMAT_REGEX_JS})#{NO_SUFFIX_REGEX}/.freeze
PROJECT_PATH_FORMAT_REGEX = /(?:#{PATH_REGEX_STR})#{NO_SUFFIX_REGEX}/.freeze
FULL_NAMESPACE_FORMAT_REGEX = %r{(#{NAMESPACE_FORMAT_REGEX}/)*#{NAMESPACE_FORMAT_REGEX}}.freeze
def root_namespace_route_regex
@root_namespace_route_regex ||= begin
illegal_words = Regexp.new(Regexp.union(TOP_LEVEL_ROUTES).source, Regexp::IGNORECASE)
single_line_regexp %r{
(?!(#{illegal_words})/)
#{NAMESPACE_FORMAT_REGEX}
}x
end
end
def full_namespace_route_regex
@full_namespace_route_regex ||= begin
illegal_words = Regexp.new(Regexp.union(ILLEGAL_GROUP_PATH_WORDS).source, Regexp::IGNORECASE)
single_line_regexp %r{
#{root_namespace_route_regex}
(?:
/
(?!#{illegal_words}/)
#{NAMESPACE_FORMAT_REGEX}
)*
}x
end
end
def project_route_regex
@project_route_regex ||= begin
illegal_words = Regexp.new(Regexp.union(ILLEGAL_PROJECT_PATH_WORDS).source, Regexp::IGNORECASE)
single_line_regexp %r{
(?!(#{illegal_words})/)
#{PROJECT_PATH_FORMAT_REGEX}
}x
end
end
def project_git_route_regex
@project_git_route_regex ||= /#{project_route_regex}\.git/.freeze
end
def root_namespace_path_regex
@root_namespace_path_regex ||= %r{\A#{root_namespace_route_regex}/\z}
end
def full_namespace_path_regex
@full_namespace_path_regex ||= %r{\A#{full_namespace_route_regex}/\z}
end
def project_path_regex
@project_path_regex ||= %r{\A#{project_route_regex}/\z}
end
def full_project_path_regex
@full_project_path_regex ||= %r{\A#{full_namespace_route_regex}/#{project_route_regex}/\z}
end
def full_namespace_format_regex
@namespace_format_regex ||= /A#{FULL_NAMESPACE_FORMAT_REGEX}\z/.freeze
end
def namespace_format_regex
@namespace_format_regex ||= /\A#{NAMESPACE_FORMAT_REGEX}\z/.freeze
end
def namespace_format_message
"can contain only letters, digits, '_', '-' and '.'. " \
"Cannot start with '-' or end in '.', '.git' or '.atom'." \
end
def project_path_format_regex
@project_path_format_regex ||= /\A#{PROJECT_PATH_FORMAT_REGEX}\z/.freeze
end
def project_path_format_message
"can contain only letters, digits, '_', '-' and '.'. " \
"Cannot start with '-', end in '.git' or end in '.atom'" \
end
def archive_formats_regex
# |zip|tar| tar.gz | tar.bz2 |
@archive_formats_regex ||= /(zip|tar|tar\.gz|tgz|gz|tar\.bz2|tbz|tbz2|tb2|bz2)/.freeze
end
def git_reference_regex
# Valid git ref regex, see:
# https://www.kernel.org/pub/software/scm/git/docs/git-check-ref-format.html
@git_reference_regex ||= single_line_regexp %r{
(?!
(?# doesn't begins with)
\/| (?# rule #6)
(?# doesn't contain)
.*(?:
[\/.]\.| (?# rule #1,3)
\/\/| (?# rule #6)
@\{| (?# rule #8)
\\ (?# rule #9)
)
)
[^\000-\040\177~^:?*\[]+ (?# rule #4-5)
(?# doesn't end with)
(?<!\.lock) (?# rule #1)
(?<![\/.]) (?# rule #6-7)
}x
end
private
def single_line_regexp(regex)
# Turns a multiline extended regexp into a single line one,
# beacuse `rake routes` breaks on multiline regexes.
Regexp.new(regex.source.gsub(/\(\?#.+?\)/ , '').gsub(/\s*/, ''), regex.options ^ Regexp::EXTENDED).freeze
end
end
end
...@@ -2,39 +2,6 @@ module Gitlab ...@@ -2,39 +2,6 @@ module Gitlab
module Regex module Regex
extend self extend self
# The namespace regex is used in Javascript to validate usernames in the "Register" form. However, Javascript
# does not support the negative lookbehind assertion (?<!) that disallows usernames ending in `.git` and `.atom`.
# Since this is a non-trivial problem to solve in Javascript (heavily complicate the regex, modify view code to
# allow non-regex validatiions, etc), `NAMESPACE_REGEX_STR_JS` serves as a Javascript-compatible version of
# `NAMESPACE_REGEX_STR`, with the negative lookbehind assertion removed. This means that the client-side validation
# will pass for usernames ending in `.atom` and `.git`, but will be caught by the server-side validation.
PATH_REGEX_STR = '[a-zA-Z0-9_\.][a-zA-Z0-9_\-\.]*'.freeze
NAMESPACE_REGEX_STR_JS = PATH_REGEX_STR + '[a-zA-Z0-9_\-]|[a-zA-Z0-9_]'.freeze
NO_SUFFIX_REGEX_STR = '(?<!\.git|\.atom)'.freeze
NAMESPACE_REGEX_STR = "(?:#{NAMESPACE_REGEX_STR_JS})#{NO_SUFFIX_REGEX_STR}".freeze
PROJECT_REGEX_STR = "(?:#{PATH_REGEX_STR})#{NO_SUFFIX_REGEX_STR}".freeze
# Same as NAMESPACE_REGEX_STR but allows `/` in the path.
# So `group/subgroup` will match this regex but not NAMESPACE_REGEX_STR
FULL_NAMESPACE_REGEX_STR = "(?:#{NAMESPACE_REGEX_STR}/)*#{NAMESPACE_REGEX_STR}".freeze
def namespace_regex
@namespace_regex ||= /\A#{NAMESPACE_REGEX_STR}\z/.freeze
end
def full_namespace_regex
@full_namespace_regex ||= %r{\A#{FULL_NAMESPACE_REGEX_STR}\z}
end
def namespace_route_regex
@namespace_route_regex ||= /#{NAMESPACE_REGEX_STR}/.freeze
end
def namespace_regex_message
"can contain only letters, digits, '_', '-' and '.'. " \
"Cannot start with '-' or end in '.', '.git' or '.atom'." \
end
def namespace_name_regex def namespace_name_regex
@namespace_name_regex ||= /\A[\p{Alnum}\p{Pd}_\. ]*\z/.freeze @namespace_name_regex ||= /\A[\p{Alnum}\p{Pd}_\. ]*\z/.freeze
end end
...@@ -52,23 +19,6 @@ module Gitlab ...@@ -52,23 +19,6 @@ module Gitlab
"It must start with letter, digit, emoji or '_'." "It must start with letter, digit, emoji or '_'."
end end
def project_path_regex
@project_path_regex ||= /\A#{PROJECT_REGEX_STR}\z/.freeze
end
def project_route_regex
@project_route_regex ||= /#{PROJECT_REGEX_STR}/.freeze
end
def project_git_route_regex
@project_route_git_regex ||= /#{PATH_REGEX_STR}\.git/.freeze
end
def project_path_regex_message
"can contain only letters, digits, '_', '-' and '.'. " \
"Cannot start with '-', end in '.git' or end in '.atom'" \
end
def file_name_regex def file_name_regex
@file_name_regex ||= /\A[[[:alnum:]]_\-\.\@\+]*\z/.freeze @file_name_regex ||= /\A[[[:alnum:]]_\-\.\@\+]*\z/.freeze
end end
...@@ -77,36 +27,8 @@ module Gitlab ...@@ -77,36 +27,8 @@ module Gitlab
"can contain only letters, digits, '_', '-', '@', '+' and '.'." "can contain only letters, digits, '_', '-', '@', '+' and '.'."
end end
def archive_formats_regex
# |zip|tar| tar.gz | tar.bz2 |
@archive_formats_regex ||= /(zip|tar|tar\.gz|tgz|gz|tar\.bz2|tbz|tbz2|tb2|bz2)/.freeze
end
def git_reference_regex
# Valid git ref regex, see:
# https://www.kernel.org/pub/software/scm/git/docs/git-check-ref-format.html
@git_reference_regex ||= %r{
(?!
(?# doesn't begins with)
\/| (?# rule #6)
(?# doesn't contain)
.*(?:
[\/.]\.| (?# rule #1,3)
\/\/| (?# rule #6)
@\{| (?# rule #8)
\\ (?# rule #9)
)
)
[^\000-\040\177~^:?*\[]+ (?# rule #4-5)
(?# doesn't end with)
(?<!\.lock) (?# rule #1)
(?<![\/.]) (?# rule #6-7)
}x.freeze
end
def container_registry_reference_regex def container_registry_reference_regex
git_reference_regex Gitlab::PathRegex.git_reference_regex
end end
## ##
......
This diff is collapsed.
...@@ -2,18 +2,6 @@ ...@@ -2,18 +2,6 @@
require 'spec_helper' require 'spec_helper'
describe Gitlab::Regex, lib: true do describe Gitlab::Regex, lib: true do
describe '.project_path_regex' do
subject { described_class.project_path_regex }
it { is_expected.to match('gitlab-ce') }
it { is_expected.to match('gitlab_git') }
it { is_expected.to match('_underscore.js') }
it { is_expected.to match('100px.com') }
it { is_expected.not_to match('?gitlab') }
it { is_expected.not_to match('git lab') }
it { is_expected.not_to match('gitlab.git') }
end
describe '.project_name_regex' do describe '.project_name_regex' do
subject { described_class.project_name_regex } subject { described_class.project_name_regex }
...@@ -44,16 +32,4 @@ describe Gitlab::Regex, lib: true do ...@@ -44,16 +32,4 @@ describe Gitlab::Regex, lib: true do
it { is_expected.not_to match('9foo') } it { is_expected.not_to match('9foo') }
it { is_expected.not_to match('foo-') } it { is_expected.not_to match('foo-') }
end end
describe '.full_namespace_regex' do
subject { described_class.full_namespace_regex }
it { is_expected.to match('gitlab.org') }
it { is_expected.to match('gitlab.org/gitlab-git') }
it { is_expected.not_to match('gitlab.org.') }
it { is_expected.not_to match('gitlab.org/') }
it { is_expected.not_to match('/gitlab.org') }
it { is_expected.not_to match('gitlab.git') }
it { is_expected.not_to match('gitlab git') }
end
end end
...@@ -37,7 +37,7 @@ describe Namespace, models: true do ...@@ -37,7 +37,7 @@ describe Namespace, models: true do
it 'rejects nested paths' do it 'rejects nested paths' do
parent = create(:group, :nested, path: 'environments') parent = create(:group, :nested, path: 'environments')
namespace = build(:project, path: 'folders', namespace: parent) namespace = build(:group, path: 'folders', parent: parent)
expect(namespace).not_to be_valid expect(namespace).not_to be_valid
end end
...@@ -238,8 +238,8 @@ describe Namespace, models: true do ...@@ -238,8 +238,8 @@ describe Namespace, models: true do
end end
context 'in sub-groups' do context 'in sub-groups' do
let(:parent) { create(:namespace, path: 'parent') } let(:parent) { create(:group, path: 'parent') }
let(:child) { create(:namespace, parent: parent, path: 'child') } let(:child) { create(:group, parent: parent, path: 'child') }
let!(:project) { create(:project_empty_repo, namespace: child) } let!(:project) { create(:project_empty_repo, namespace: child) }
let(:path_in_dir) { File.join(repository_storage_path, 'parent', 'child') } let(:path_in_dir) { File.join(repository_storage_path, 'parent', 'child') }
let(:deleted_path) { File.join('parent', "child+#{child.id}+deleted") } let(:deleted_path) { File.join('parent', "child+#{child.id}+deleted") }
......
...@@ -100,7 +100,7 @@ describe KubernetesService, models: true, caching: true do ...@@ -100,7 +100,7 @@ describe KubernetesService, models: true, caching: true do
it 'sets the namespace to the default' do it 'sets the namespace to the default' do
expect(kube_namespace).not_to be_nil expect(kube_namespace).not_to be_nil
expect(kube_namespace[:placeholder]).to match(/\A#{Gitlab::Regex::PATH_REGEX_STR}-\d+\z/) expect(kube_namespace[:placeholder]).to match(/\A#{Gitlab::PathRegex::PATH_REGEX_STR}-\d+\z/)
end end
end end
end end
...@@ -187,7 +187,7 @@ describe KubernetesService, models: true, caching: true do ...@@ -187,7 +187,7 @@ describe KubernetesService, models: true, caching: true do
kube_namespace = subject.predefined_variables.find { |h| h[:key] == 'KUBE_NAMESPACE' } kube_namespace = subject.predefined_variables.find { |h| h[:key] == 'KUBE_NAMESPACE' }
expect(kube_namespace).not_to be_nil expect(kube_namespace).not_to be_nil
expect(kube_namespace[:value]).to match(/\A#{Gitlab::Regex::PATH_REGEX_STR}-\d+\z/) expect(kube_namespace[:value]).to match(/\A#{Gitlab::PathRegex::PATH_REGEX_STR}-\d+\z/)
end end
end end
end end
......
...@@ -287,7 +287,7 @@ describe API::Users do ...@@ -287,7 +287,7 @@ describe API::Users do
expect(json_response['message']['projects_limit']). expect(json_response['message']['projects_limit']).
to eq(['must be greater than or equal to 0']) to eq(['must be greater than or equal to 0'])
expect(json_response['message']['username']). expect(json_response['message']['username']).
to eq([Gitlab::Regex.namespace_regex_message]) to eq([Gitlab::PathRegex.namespace_format_message])
end end
it "is not available for non admin users" do it "is not available for non admin users" do
...@@ -459,7 +459,7 @@ describe API::Users do ...@@ -459,7 +459,7 @@ describe API::Users do
expect(json_response['message']['projects_limit']). expect(json_response['message']['projects_limit']).
to eq(['must be greater than or equal to 0']) to eq(['must be greater than or equal to 0'])
expect(json_response['message']['username']). expect(json_response['message']['username']).
to eq([Gitlab::Regex.namespace_regex_message]) to eq([Gitlab::PathRegex.namespace_format_message])
end end
it 'returns 400 if provider is missing for identity update' do it 'returns 400 if provider is missing for identity update' do
......
...@@ -462,6 +462,8 @@ describe 'project routing' do ...@@ -462,6 +462,8 @@ describe 'project routing' do
expect(get('/gitlab/gitlabhq/blob/master/app/models/compare.rb')).to route_to('projects/blob#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'master/app/models/compare.rb') expect(get('/gitlab/gitlabhq/blob/master/app/models/compare.rb')).to route_to('projects/blob#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'master/app/models/compare.rb')
expect(get('/gitlab/gitlabhq/blob/master/app/models/diff.js')).to route_to('projects/blob#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'master/app/models/diff.js') expect(get('/gitlab/gitlabhq/blob/master/app/models/diff.js')).to route_to('projects/blob#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'master/app/models/diff.js')
expect(get('/gitlab/gitlabhq/blob/master/files.scss')).to route_to('projects/blob#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'master/files.scss') expect(get('/gitlab/gitlabhq/blob/master/files.scss')).to route_to('projects/blob#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'master/files.scss')
expect(get('/gitlab/gitlabhq/blob/master/blob/index.js')).to route_to('projects/blob#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'master/blob/index.js')
expect(get('/gitlab/gitlabhq/blob/blob/master/blob/index.js')).to route_to('projects/blob#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'blob/master/blob/index.js')
end end
end end
...@@ -470,6 +472,8 @@ describe 'project routing' do ...@@ -470,6 +472,8 @@ describe 'project routing' do
it 'to #show' do it 'to #show' do
expect(get('/gitlab/gitlabhq/tree/master/app/models/project.rb')).to route_to('projects/tree#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'master/app/models/project.rb') expect(get('/gitlab/gitlabhq/tree/master/app/models/project.rb')).to route_to('projects/tree#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'master/app/models/project.rb')
expect(get('/gitlab/gitlabhq/tree/master/files.scss')).to route_to('projects/tree#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'master/files.scss') expect(get('/gitlab/gitlabhq/tree/master/files.scss')).to route_to('projects/tree#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'master/files.scss')
expect(get('/gitlab/gitlabhq/tree/master/tree/files')).to route_to('projects/tree#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'master/tree/files')
expect(get('/gitlab/gitlabhq/tree/tree/master/tree/files')).to route_to('projects/tree#show', namespace_id: 'gitlab', project_id: 'gitlabhq', id: 'tree/master/tree/files')
end end
end end
......
...@@ -3,246 +3,46 @@ require 'spec_helper' ...@@ -3,246 +3,46 @@ require 'spec_helper'
describe DynamicPathValidator do describe DynamicPathValidator do
let(:validator) { described_class.new(attributes: [:path]) } let(:validator) { described_class.new(attributes: [:path]) }
# Pass in a full path to remove the format segment: describe '#path_valid_for_record?' do
# `/ci/lint(.:format)` -> `/ci/lint` context 'for project' do
def without_format(path) it 'calls valid_project_path?' do
path.split('(', 2)[0] project = build(:project, path: 'activity')
end
# Pass in a full path and get the last segment before a wildcard
# That's not a parameter
# `/*namespace_id/:project_id/builds/artifacts/*ref_name_and_path`
# -> 'builds/artifacts'
def path_before_wildcard(path)
path = path.gsub(STARTING_WITH_NAMESPACE, "")
path_segments = path.split('/').reject(&:empty?)
wildcard_index = path_segments.index { |segment| parameter?(segment) }
segments_before_wildcard = path_segments[0..wildcard_index - 1]
segments_before_wildcard.join('/')
end
def parameter?(segment)
segment =~ /[*:]/
end
# If the path is reserved. Then no conflicting paths can# be created for any
# route using this reserved word.
#
# Both `builds/artifacts` & `build` are covered by reserving the word
# `build`
def wildcards_include?(path)
described_class::WILDCARD_ROUTES.include?(path) ||
described_class::WILDCARD_ROUTES.include?(path.split('/').first)
end
def failure_message(missing_words, constant_name, migration_helper)
missing_words = Array(missing_words)
<<-MSG
Found new routes that could cause conflicts with existing namespaced routes
for groups or projects.
Add <#{missing_words.join(', ')}> to `DynamicPathValidator::#{constant_name} expect(described_class).to receive(:valid_project_path?).with(project.full_path).and_call_original
to make sure no projects or namespaces can be created with those paths.
To rename any existing records with those paths you can use the expect(validator.path_valid_for_record?(project, 'activity')).to be_truthy
`Gitlab::Database::RenameReservedpathsMigration::<VERSION>.#{migration_helper}`
migration helper.
Make sure to make a note of the renamed records in the release blog post.
MSG
end
let(:all_routes) do
Rails.application.routes.routes.routes.
map { |r| r.path.spec.to_s }
end
let(:routes_without_format) { all_routes.map { |path| without_format(path) } }
# Routes not starting with `/:` or `/*`
# all routes not starting with a param
let(:routes_not_starting_in_wildcard) { routes_without_format.select { |p| p !~ %r{^/[:*]} } }
let(:top_level_words) do
routes_not_starting_in_wildcard.map do |route|
route.split('/')[1]
end.compact.uniq
end
# All routes that start with a namespaced path, that have 1 or more
# path-segments before having another wildcard parameter.
# - Starting with paths:
# - `/*namespace_id/:project_id/`
# - `/*namespace_id/:id/`
# - Followed by one or more path-parts not starting with `:` or `*`
# - Followed by a path-part that includes a wildcard parameter `*`
# At the time of writing these routes match: http://rubular.com/r/Rv2pDE5Dvw
STARTING_WITH_NAMESPACE = %r{^/\*namespace_id/:(project_)?id}
NON_PARAM_PARTS = %r{[^:*][a-z\-_/]*}
ANY_OTHER_PATH_PART = %r{[a-z\-_/:]*}
WILDCARD_SEGMENT = %r{\*}
let(:namespaced_wildcard_routes) do
routes_without_format.select do |p|
p =~ %r{#{STARTING_WITH_NAMESPACE}/#{NON_PARAM_PARTS}/#{ANY_OTHER_PATH_PART}#{WILDCARD_SEGMENT}}
end
end
# This will return all paths that are used in a namespaced route
# before another wildcard path:
#
# /*namespace_id/:project_id/builds/artifacts/*ref_name_and_path
# /*namespace_id/:project_id/info/lfs/objects/*oid
# /*namespace_id/:project_id/commits/*id
# /*namespace_id/:project_id/builds/:build_id/artifacts/file/*path
# -> ['builds/artifacts', 'info/lfs/objects', 'commits', 'artifacts/file']
let(:all_wildcard_paths) do
namespaced_wildcard_routes.map do |route|
path_before_wildcard(route)
end.uniq
end
STARTING_WITH_GROUP = %r{^/groups/\*(group_)?id/}
let(:group_routes) do
routes_without_format.select do |path|
path =~ STARTING_WITH_GROUP
end
end
let(:paths_after_group_id) do
group_routes.map do |route|
route.gsub(STARTING_WITH_GROUP, '').split('/').first
end.uniq
end
describe 'TOP_LEVEL_ROUTES' do
it 'includes all the top level namespaces' do
failure_block = lambda do
missing_words = top_level_words - described_class::TOP_LEVEL_ROUTES
failure_message(missing_words, 'TOP_LEVEL_ROUTES', 'rename_root_paths')
end end
expect(described_class::TOP_LEVEL_ROUTES)
.to include(*top_level_words), failure_block
end end
end
describe 'GROUP_ROUTES' do context 'for group' do
it "don't contain a second wildcard" do it 'calls valid_group_path?' do
failure_block = lambda do group = build(:group, :nested, path: 'activity')
missing_words = paths_after_group_id - described_class::GROUP_ROUTES
failure_message(missing_words, 'GROUP_ROUTES', 'rename_child_paths')
end
expect(described_class::GROUP_ROUTES) expect(described_class).to receive(:valid_group_path?).with(group.full_path).and_call_original
.to include(*paths_after_group_id), failure_block
end
end
describe 'WILDCARD_ROUTES' do expect(validator.path_valid_for_record?(group, 'activity')).to be_falsey
it 'includes all paths that can be used after a namespace/project path' do
aggregate_failures do
all_wildcard_paths.each do |path|
expect(wildcards_include?(path))
.to be(true), failure_message(path, 'WILDCARD_ROUTES', 'rename_wildcard_paths')
end
end end
end end
end
describe '.without_reserved_wildcard_paths_regex' do context 'for user' do
subject { described_class.without_reserved_wildcard_paths_regex } it 'calls valid_user_path?' do
user = build(:user, username: 'activity')
it 'rejects paths starting with a reserved top level' do expect(described_class).to receive(:valid_user_path?).with(user.full_path).and_call_original
expect(subject).not_to match('dashboard/hello/world')
expect(subject).not_to match('dashboard')
end
it 'matches valid paths with a toplevel word in a different place' do expect(validator.path_valid_for_record?(user, 'activity')).to be_truthy
expect(subject).to match('parent/dashboard/project-path') end
end
it 'rejects paths containing a wildcard reserved word' do
expect(subject).not_to match('hello/edit')
expect(subject).not_to match('hello/edit/in-the-middle')
expect(subject).not_to match('foo/bar1/refs/master/logs_tree')
end
it 'matches valid paths' do
expect(subject).to match('parent/child/project-path')
end
end
describe '.regex_excluding_child_paths' do
let(:subject) { described_class.without_reserved_child_paths_regex }
it 'rejects paths containing a child reserved word' do
expect(subject).not_to match('hello/group_members')
expect(subject).not_to match('hello/activity/in-the-middle')
expect(subject).not_to match('foo/bar1/refs/master/logs_tree')
end
it 'allows a child path on the top level' do
expect(subject).to match('activity/foo')
expect(subject).to match('avatar')
end
end
describe ".valid?" do
it 'is not case sensitive' do
expect(described_class.valid?("Users")).to be_falsey
end
it "isn't valid when the top level is reserved" do
test_path = 'u/should-be-a/reserved-word'
expect(described_class.valid?(test_path)).to be_falsey
end
it "isn't valid if any of the path segments is reserved" do
test_path = 'the-wildcard/wikis/is-not-allowed'
expect(described_class.valid?(test_path)).to be_falsey
end
it "is valid if the path doesn't contain reserved words" do
test_path = 'there-are/no-wildcards/in-this-path'
expect(described_class.valid?(test_path)).to be_truthy
end
it 'allows allows a child path on the last spot' do
test_path = 'there/can-be-a/project-called/labels'
expect(described_class.valid?(test_path)).to be_truthy
end
it 'rejects a child path somewhere else' do
test_path = 'there/can-be-no/labels/group'
expect(described_class.valid?(test_path)).to be_falsey
end end
it 'rejects paths that are in an incorrect format' do context 'for user namespace' do
test_path = 'incorrect/format.git' it 'calls valid_user_path?' do
user = create(:user, username: 'activity')
expect(described_class.valid?(test_path)).to be_falsey namespace = user.namespace
end
end
describe '#path_reserved_for_record?' do expect(described_class).to receive(:valid_user_path?).with(namespace.full_path).and_call_original
it 'reserves a sub-group named activity' do
group = build(:group, :nested, path: 'activity')
expect(validator.path_reserved_for_record?(group, 'activity')).to be_truthy expect(validator.path_valid_for_record?(namespace, 'activity')).to be_truthy
end end
it "doesn't reserve a project called activity" do
project = build(:project, path: 'activity')
expect(validator.path_reserved_for_record?(project, 'activity')).to be_falsey
end end
end end
...@@ -252,7 +52,7 @@ describe DynamicPathValidator do ...@@ -252,7 +52,7 @@ describe DynamicPathValidator do
validator.validate_each(group, :path, "Path with spaces, and comma's!") validator.validate_each(group, :path, "Path with spaces, and comma's!")
expect(group.errors[:path]).to include(Gitlab::Regex.namespace_regex_message) expect(group.errors[:path]).to include(Gitlab::PathRegex.namespace_format_message)
end end
it 'adds a message when the path is not in the correct format' do it 'adds a message when the path is not in the correct format' do
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment