sf-squash-news Add crystal to expansion keys

Smalltalk support

Ensure `dist` and `group` are set in job config

Back out random spec ordering

allow api requests no matter if pushes/pull_requests are enabled in settings

use progress rspec format

turn rspec format back to documentation

Check for :group || 'group' instead of :group || 'dist'

Add specs for Dist and Group normalizers

Fix a few rubocop offenses

add migration for creating starred_repositories table

remove unnecessary column definitions

add index for user and repo ids on starred_repos

fix syntax of drop statement

update index migration

change name of starred_repositories table to stars

fix typo

Default dist to trusty if `docker` is in services

fix bad migration dates in filename

Update 20151112153500_create_starred_repositories.rb

Correct `repo_id` => `repository_id` in migration

Add info about default branch to repository and build for pusher

Set dist per matrix include entry

Guard against bogus config and matrix values in dist normalization

Also skip non-hashy matrix values when normalizing dist

Apply same `config_hashy?` test to matrix.include entries

Update 20151112153500_create_stars.rb

Update 20151113111400_stars_add_index_user_id_and_repo_id.rb

Update 20151113111400_stars_add_index_user_id_and_repo_id.rb

Rename 20151113111400_stars_add_index_user_id_and_repo_id.rb to 20151113111400_stars_add_index_user_id_and_repository_id.rb

fix model name syntax

Revert "Merge pull request #493 from travis-ci/send-default-branch-info-in-pusher"

We've started seeing performance issues in hub after deploying this
change. I'm going to revert that here as well just in case this
happens in gatekeeper too.

This reverts commit 3c20e66a20d9d540e2a6c5fb70c542fe97c62e25, reversing
changes made to c7ea1bb326b32898df5568e2699466cdaf5395b3.

Add default branch info to repository in builds payload for pusher

add indexes to reduce sequetial scans

fix table for index onannotations

Consider API builds for repo's build statuses

When determining a repo's statuses and last_build on a given branch,
API builds should also be considered.

Restructure #last_build_on specs

Restructure #last_build specs

Restructure #last_build and #last_build_on specs

add first-logged-in to users table

add first_logged_in to attr_accessible

fix attribute name

Add logging around job queue selection/matching

Remove nonexistent job id ref

Log in Queue.for when default_queue is selected

Allow to limit repos list

check for current_user during run and accept?

Handle empty matrix.include

Do not throw NoMethodError when matrix.include is empty

Optimize query that is done for clearing log parts

parts.delete_all will first fetch every log part and then do a delete
query for each log part that is retrieved, while
Log::Part.where().delete_all will run a single query to delete every log
part. delete_all does mean that the callbacks are bypassed, but that was
the same as the previous behaviour.

Reload log parts after deletion

That way `log.parts.length` will reload, causing the tests to pass, and
other things that use `log.parts` won't break.

Reject build request if compare URL is too long

The type for compare URLs in our database is VARCHAR(255).
Rather than accepting the request and throwing an exception
later (without a chance to let users know), we reject the
build request and advise to shorten branch/tag name.

Stub commit object to include compare_url

reduce complexity of updating branch.last_build_id

only update the branch for push and api requests

Add R language key

add receive class for cron

Update db/migrate

To correspond to
https://github.com/travis-ci/travis-migrations/commit/fcf6eea3e3122a7cbb857826db835de69974c54d

This allows specs in travis-api to see job.debug_options

remove db folder, update rakefile and .travis.yml

update rakefile

create rake version dependency to fix missing method last_comment

restore require travis

remove duplicate spec run

add statement_timeout to db config

Add ability to get running builds in API v2

remove Jruby from .travis.yml

Create shared parent class for logs classes

Database connections in ActiveRecord are bound to a class, so in order
to use only one connection for both classes we need to have a shared
parent class.

tidy Rakefile

chenge rake to rspec

update readme

Update README.md

make rate_limit setting accessible and add validation

remove blank space

fix method name

add errors attribute to repo

remove errors attribute

set default value for rate_limit if nil

add spec to check validation of rate_limit

renaming

Rename rate_limit to api_builds_rate_limit

rearrange settings config hash, add test for default api build rate limit

Return nil if no value is set for api_builds_rate_limit

relax hashr gem requirement

backport travis-config v1.0.x resource url loading

fixes after updating travis-config

allow TRAVIS_* resource urls

add specs for TRAVIS_* resource urls

allow [TRAVIS_]LOGS_DATABASE_URL

Remove duplicate in config and add oauth2

Remove all the duplicates in the rspec tests

Remove a comma :)

Coerce config into a Hash before passing it to S3 service

Hashr does not appear to implement `#slice`, and we are getting
`{}` when we call it.

Remove pry

Add active to pusher's repository payload on build events

Downcase retrieved language on config before creating the build@

clarify documentation

Revert " Downcase retrieved language on config before creating the build"

Add google-api-client

Also free up thor version requirement, which appears unnecessary.

Roll caches into a method

Remove intermediate `#bucket` call

Buckets are easy to work with in S3, but for GCS, this is not the
case.
For example, it is not quite easy (if not impossible) to
get a list of objects belonging to the bucket.
Instead, one has to do:

storage.list_objects(bucket_name, prefix: prefix).items

(This part of GCS logic is not yet implemented.)

Preliminary implementation of GCS cache listing

There are now two wrapper classes for the cache
objects, backed by S3 and GCS.

S3 credentials warning is removed for the time being

Avoid bucket being nil

Find Google objects correctly

Actually require googel api library

Add GcsWrapper objects to list of caches

Add repository to GcsWrapper

Memoize @caches

Clean up debugging output

Ensure GCS's items map is nonempty

Minor formatting changes

Define `#content` to fetch cache's content

So that admin can stream it to its client

Download GCS cache contents to memory

Rewind GCS cache content before loading to memory

Reinstate setup? check

Instead of instantiating and checking whether
any cache exists, we just simply look at the configuration
parameter cache_options.

If there is no cache to begin with, we will be returning `[]`
anyway, so this change is not a bad thing.

Add bare minimum specs for GCS caching

Downcase language attribute before saving the new build

Add comment to before_create method

expand 'smalltalk_config' key

Set language before expanding matrix

Refactor cache_options validation code

Downcase language directly on the Language class

This modifies the way we set config[:language] on the Language class, in order to downcase it or use the DEFAULT_LANGUAGE.
This way, the language modification is propagated to the database, accessible everywhere, and mainly the language attribute on the .travis.yml file won't be case sensitive anymore.

Add missing GCS support file

Add `#source` to cache wrapper object

This allows easy distinction by consumers of the `find_caches`
service where the cache resides

Ensure allow_failures config is an array

Ensure config[:addons] is reasonable when deploy is present

We should drop non-Hash addons config when a top-level
deploy key is present.

When the top-level deploy key is present, while normalizing,
an addons value that is not a Hash will be sent `#[]`,
which is raises `NoMethodError`.

Remove superflulous initialization

Since we handle the non-Hash case of `config[:addons]`,
this is no longer necessary

Ensure `#commit` does not throw NoMethodError

Under certain circumstances the payload from GitHub could lack
data we assume to exist, leading to NoMethodError

Prefer 'missing config' error message to 'github pages branch'

Fixes https://github.com/travis-ci/travis-ci/issues/5933

If `.travis.yml` contains a YAML syntax error on the `gh-pages`
branch, we display the error message "github pages branch" even if
`gh-pages` is (textually) whitelisted in that file.

We should display "missing config" instead.

Fix gh-pages message spec

The config should be valid in this spec

Tweak 'config missig' error message

Config can be blank if exception is thrown while processing the file
due to YAML syntax error

Add memcached service to build

Pin listen to < 3.1.0

3.1.0 requires Ruby 2.2. https://rubygems.org/gems/listen/versions/3.1.0

Pin listen to < 3.1.0

3.1.0 requires Ruby 2.2. https://rubygems.org/gems/listen/versions/3.1.0

change older_than builds to order by number

fix order by section

Fix StateCache spec

Test data have been written, but not modified, so the value
should remain 'failed'

Check current sent notifications

Add canceled event to handler

Move config.blank? to the end

Add current_build_id to pusher

Jonne Haß authored Sven Fuchs committed

Push event #4723 passed

  • Ran for
  • Total time
  • New branch build