- Project groups
- External IDs for build configuration and template
- Project configuration files changes
- Meta runner
- Queued build page
- Health status report
- Branch filters in triggers
- Mercurial subrepo improvements
Several important improvements and bug fixes have been made in project groups feature.
- "Configure Visible Projects" dialog has become aware of project hierarchy
- projects hierarchy is shown on Administration -> Projects page the same way as it shown in the Projects popup
- projects hierarchy is shown on Administration -> Disk usage page
- projects hierarchy is shown in all drop downs where a list of projects or configurations is displayed
- investigations page for a project also shows all investigations from sub-projects. The same applies to muted problems page, current problems, build chains and project change log. All these pages will show sub-projects too.
- notification rule defined for a project will be effective for sub-projects too
- builds schedule page for a project shows schedule from sub-projects too
- it is now possible to define cleanup rules for projects, such cleanup rules serve as default for configurations from this project and sub project. Default cleanup rule is replaced with cleanup rule in Root project.
VCS roots and templates
There is no notion of shared VCS root anymore. If VCS root should be available to several projects it should be moved to common parent of these projects or to Root project. Thus when you upgrade to this EAP build, all shared VCS roots will be moved to Root project.
The same applies to templates, build configuration can be attached to a template if template belongs to project of this configuration or to one of parent projects. This means that starting with this EAP you won't be able to use template from separate project hierarchy. However, as this was possible before, TeamCity will still support this case and will load such configurations without errors.
External IDs for build configuration and template
It is now possible to define external ID for build configuration or template. It works the same way as external ID for project, i.e. it is used in URLs instead of internal ID, and it is used in configuration files. However with build configuration external IDs add some additional benefits:
- if you're using dependency parameters, you can use external IDs there instead of cryptic btXXX:
dep.<my build configuration external id>.build.number
- you can access build artifacts using a URL like this:
http://<host>/repository/download/<my build configuration external id>/lastSuccessful/artifact.zip
- external ID can be used when build is triggered by HTTP request and in REST requests
Project configuration files changes
We continue refactoring the way how configuration files are stored on disk under TeamCity data directory. This EAP brings two important changes:
- from now on, build configuration internal ID is no longer stored in project configuration files, external ID is used instead, the same applies to build configuration template.
- since all VCS roots now belong to projects we removed the global
vcs-roots.xmlfile and moved VCS roots settings to corresponding projects.
Meta-runner is a Build Runner acting as a combination of one or more build runners and having user interface targeted to user domain. For example, you can define one or more build steps in some build configuration to solve some specific task. Now if you need to reuse this set of steps, you should generalize them, i.e. replace some hard coded values with parameter references and create a template with this set of steps. Then you can reuse this template in several configurations. However, due to template nature this approach is still not flexible enough. Template brings a lot more settings in your configuration, and many of them you might not want to reuse. It would be great to be able to reuse this set of build steps only, preferably as a dedicated runner with own web interface, and this is where Meta-runner comes into play.
Let's see how it works. Sometimes during the build process we need to replace some pattern in several files before making a distribution package. For example, if you build a TeamCity plugin you need to provide build number in teamcity-plugin.xml file bundled with the plugin. So this build runner should have the following parameters:
- pattern to search in files
- replacement string
- patterns for files where replacement must be performed
First of all I need to create a build configuration with Ant build step. I will use Ant built-in replaceregexp task to do all the work, so my build.xml will look like this:
Note that in several places I used parameter references:
- %pattern_to_replace% - regexp pattern to search in files
- %substitution% - substitution string
- %basedir% - base directory where to perform replacement
- %files_to_include% and %files_to_exclude% - Ant include / exclude patterns for files
In the build configuration where this Ant build step is defined I will see these parameter references as undefined parameters on Build parameters tab, because no one provided values for them.
Now we need to associate labels and descriptions with these parameters, this can be done through regular parameter editing dialog.
Once we've described parameters we can extract meta runner from the build steps page of the build configuration:
At this point we can also add requirements in configuration. This is useful If our runner depends on some tools, or software installed on agent. If added these requirements will be extracted to Meta runner too. But in this example, our runner is generic and can run everywhere.
And use it in some other build configuration:
Definitions of meta runners are stored under
TeamCity Data Directory/config/_meta_runners/ directory. Here is complete definition for Replace in Files build runner (I added validation regexp for parameter values and reordered parameters):
In order to install this build runner in your TeamCity installation you need to save this definition to a file under
TeamCity Data Directory/config/_meta_runners/ directory. File should have name like:
<runner id>.xml, where <runner id> is unique identifier of this build runner. Server will detect this definition and will load it on the fly.
- when meta-runner is extracted from web interface there is no way to specify which steps you want to extract, all defined build steps will be extracted
- extract meta-runner interface does not allow to reorder parameters, you have to open extracted file and change order manually
- meta runner definitions are global for the server, there is no way to store them in some project only
Queued build page
Dependencies progress has been added on queued build page, you can now see estimates for all dependencies in one place.
Health status report
Huge logs (based on disk usage report)
This report shows builds with huge log files. Huge (hundreds of megabytes) build logs are rarely useful, it is hard to analyze them and in most of cases they just waste space on disk. We hope this report will help to find configurations that produce such log files and fix them.
Swabra clean checkout
This report finds incorrect settings of Swabra build feature which can cause unnecessary clean checkout.
Branch filters in triggers
VCS, Schedule and Finish build triggers have got new setting: Branch filter. With the help of the new filter you can limit the set of branches where automatic triggering will be performed. Read more about branch filters in our documentation: Working with Feature Branches#Triggers
Mercurial subrepo improvements
We continue improving usability of Mercurial subrepositories changes. To be able to distinguish main repository changes and subrepository changes changes from subrepositories now have additional icon:
Note: on this screenshot you can also see another feature: dashed lines from change in main repository to changes in subrepositories. These dashed lines are shown if you click on the change in main repository in commits graph.
One more improvement in graph of commits allows to collapse VCS root commits into single line:
- Server ID - to be able to remove duplicates from our usage statistics reports we need to identify server somehow. To be able to do it we introduced a unique identifier - server ID, which is sent together with usage statistics report. This ID is completely anonymous and does not contain machine-specific information.
- Schedule trigger now supports defining time to trigger in a specific timezone.
- Builds schedule report allows to filter data by sub projects
- Build problems UI has got several improvements, notifications for build problems investigations are now supported