This task fails because somewhere deep in $(Build.SourcesDirectory)/... is a broken symlink.

- task: CopyFiles@2
    inputs:
      sourceFolder: $(Build.SourcesDirectory)
      contents: 'MLO'
      targetFolder: $(Build.ArtifactStagingDirectory)

Unhandled: Failed find: ENOENT: no such file or directory, stat '/__w/1/s/arm-buildroot-linux-gnueabihf_sdk-buildroot/arm-buildroot-linux-gnueabihf/sysroot/bin/systemd-resolve'

I would not expect this task to fail on broken symlink but even less in this case because my content pattern resolves to a single file.

Anybody else think this is not the correct behavior?

How should I work around this given I cannot get rid of the symlink?

In the visual editor, there is a checkbox:

This checkbox adds pwsh: true to the yml when you hit the 'View YAML' button:

steps:
- powershell: 'Install-Module platyPS -Force -Confirm:$false -Scope CurrentUser'
  pwsh: true
  displayName: 'Install-Module platyPS'

However, if you put this into a yml file, when attempting to run the build on a Windows worker (VS2017), this error occurs in the pipeline:

Is there some other way to identify the use of pwsh (PowerShell Core 6) in this task in the yml?

copy from: microsoft/azure-pipelines-agent#1887

Have you tried trouble shooting?
N/A

Agent Version and Platform
Agent version: 2.140.2

OS of the machine running the agent? OSX/Windows/Linux/...
Windows

VSTS Type and Version
VisualStudio.com

If VisualStudio.com, what is your account name?
http://uipath.visualstudio.com

What's not working?
As per this issue, we know that some time ago the agent supported running jobs locally (it was some kind of a hidden feature, but this prospect excited us). What's the status on this? Is there any way of running builds locally? This would really help for speeding up testing!

Thank you!

Hello,

I'm trying to define a variable to be used in later phases. It seems the way you access variables $(varName) does not work inside of if statements.

Here's what I have:

variables:
Build.Repository.Clean: true
_enableTelemetry: true
_HelixType: build/product

buildType: public
${{ if eq(variables.System.TeamProject, 'internal') }}:
${{ if notin(variables.Build.Reason, 'PullRequest') }}:
${{ if not(contains(variables.Build.DefinitionName, 'github')) }}:
buildType: internal

I've trying to branch based on the value of buildType later, but I'm having trouble doing so.

I've tried all of the following:

${{ if eq('$(buildType)', 'public') }}:
${{ if eq($(buildType), 'public') }}:
${{ if eq(variables.buildType, 'public') }}:
${{ if eq('variables.buildType', 'public') }}:
${{ if eq(variables['buildType'], 'public') }}:

But nothing seems to work.

Does this not work in yaml or am I just doing it wrong?

Thanks!
-Adam

My understanding is that the name of a counter can be set to a variable and that this variable is then expanded to be the counter name. Then, when the variable value changes, the counter re-seeds. Is this correct?

If so, is it supported when the variable is from a variable group?

I have this yaml:

variables:
  - group: Build_Identifiers
  - name: BUILD_NUMBER
    value: $[ counter( variables['majorMinorVersion'], 1) ]

majorMinorVersion is from the Build_Identifiers variable group. When the value of majorMinorVersion is changed, the counter does not reset to 1.

Apologies if issues are reserved for template issues.

I am struggling to make any progress towards getting a package R CMD checked on a Win VM. Any tips would be appreciated.

It seems I cannot run a UniversalPackages@0 task in a container job unless I add a bunch of dependencies to my container image.

I would assume the task to be self sufficient, no?

As an example, when trying to download a package I get:

Error: An unexpected error occurred while trying to download the package. Exit code(136) and error(Failed to load ؖ!�, error: libunwind.so.8: cannot open shared object file: No such file or directory
Failed to bind to CoreCLR at '/__t/ArtifactTool/0.2.88/x64/libcoreclr.so')
Packages failed to download

I have a need to use the minimatch file pattern functionality in PowerShell. In the VSTS Task SDK there is the Find-Match function that implements this. How can I use this (or equivalent) in a YAML PowerShell step so I can use pattern matching outside of the built in tasks?

Note that I have need for this in a step template so simply copying the .ps1 implementation file from GitHub wouldn't work because templates don't work with external .ps1 files at this time.

I've noticed, that if I create a new project in Azure DevOps and initialize it using my existing local git repo which includes '.vsts-ci.yml' file, a build definition doesn't get created automatically and in order to trigger one to be created I need to make some changes to the file and push those changes up to the origin.

The thing is that I have the '.vsts-ci.yml' file inside of my project template and I'd really like it to automatically generate the build definition straight on the initial push (so that I could then go and configure branch policies for the master to depend on this build).

Is this an expected behavior or it's actually a bug?

Thanks!

Hello,

We have a lot of ${{ if .... }}: in our yaml, and it would be nice to be able to use else as well instead of having to manually write out the inverse of the condition every time.

Any plans for doing this?

Thanks,
-Adam

Please add schedule triggers to yaml (currently not supported).

Here pipelines is displayed as "denoland.deno_std"

Is it possible now to join the elements of sequence to a string?

items:
- foo
- bar

result: ${{ join(items, ';') }}

where result would contain foo;bar string

One of possible scenarios for this could specifying MSBuild semicolon-separated property values using array syntax (instead of a very long string)

The following works nicely, except if the script length is larger than 2000 chars. Some of us need such long scripts :-)

    - bash: |
        if [ '${{parameters.some_param}}' == true ]; then
          do_something
        fi

With larger scripts, the build fails at the template substitution step with the following message:

azure-pipelines.yml (Line: xxx, Col: yyy): Exceeded max expression length 2000. For more help, refer to https://go.microsoft.com/fwlink/?linkid=842996

Looking at this document (which I found with some help) it seems that this repository contains all the necessary functionality to validate YAML pipeline definitions locally, without actually sending any build commands to build servers.

I tried to use run.sh --yaml --what-if as shown in that document, but that complained about not finding bin/Agent.Listener and I couldn't find any instructions on how to build that file correctly. However, based on the descriptions in the document, it seems to do exactly what I would need to avoid having to push code to the repository just to see if my build config is valid.

It would be awesome if that could be exposed as something I can run from within my repository (e.g. a dotnet global tool).

Hey, I have a central step template, which I use in different builds. One build started to freakout, and does only part of one step of the steps. No changes were done to yaml in the past 3 weeks. How do I fix this? Running with debug shows no errors. It looks like it just skips some of the commands.

Is it possible to nest templates? My first attempt at doing this is throwing an error of android-modules.yml (Line: 1, Col: 1): Unexpected value 'resources' within the top level phase template. This seems to be not a possible option then. Would it be better to use conditions within the phase template to decide whether or not to execute a step?

Thanks

Hello,
I'm trying to commit a file that is created during build back into the git repo. I've followed the instructions on this page to enable rights for Project Collection Build Service and I also made sure that persistCredentials is set to true. When I do git config --list in the PowerShell script I can see the line with http.https:(someurl)=AUTHORIZATION: bearer ***, which was set as the last step in checkout. However, I'm getting the error below. Do I need to setup the user email (what would that be with respect to the Azure DevOps project?) and the password as the access token or is there something else that I'm missing? Thank you!

pool:
  vmImage: 'VS2017-Win2016'

variables:
  solution: '**/*.sln'
  buildPlatform: 'Any CPU'
  buildConfiguration: 'Release'

steps:
- checkout: self
  persistCredentials: true
  clean: true

- powershell: |
   git --version
   New-Item -Path "$(System.DefaultWorkingDirectory)" -Name "testfile1.txt" -ItemType "file" -Value "This is a text string."
   Write-Host "new file created, now adding file to git"
   git -c http.extraheader="AUTHORIZATION: bearer $(System.AccessToken)" add -A
   Write-Host "git commit with message"
   git -c http.extraheader="AUTHORIZATION: bearer $(System.AccessToken)" commit -a -m "Test Commit from Azure DevOps"
   Write-Host "git push"
   git -c http.extraheader="AUTHORIZATION: bearer $(System.AccessToken)" push

The YAML schema described on this repo here differs from the YAMl schema on the Microsoft site here. Which is it?

From Microsoft site

name: string  # build numbering format
resources:
  containers: [ containerResource ]
  repositories: [ repositoryResource ]
variables: { string: string } | [ variable ]
trigger: trigger
pr: pr
jobs: [ job | templateReference ]

From this repo

resources:
  pipelines: [ pipeline ]  
  builds: [ build ]
  repositories: [ repository ]
  containers: [ container ]
  packages: [ package ]

I've found that "pipelines" isn't recognized on Azure Pipelines

I can't find the docs for Yaml for authorization. These docs used to live at:

https://github.com/Microsoft/azure-pipelines-agent/blob/3d39fca3cac2bf54c16600e0a3b9e7972d4c5ef3/docs/preview/yamlgettingstarted-authz.md

Now they are gone, as you can see here:

https://github.com/Microsoft/azure-pipelines-agent/blob/master/docs/preview/yamlgettingstarted.md

I couldn't find the information that was available on those md files anywhere else. Did I miss it somewhere on azure-pipelines-agent or on this repo or is it really missing?

Is it possible to regex a build variable?

ex: $(System.TeamProject) = TEAM1-TestProject

I would like get anything to the left of the hyphen 'TEAM1'. Is this possible?

It would be great to have an example about how to publish an npm module to the public npmjs.com registry from a github repository. Ideally, this would illustrate how to trigger a new CI build on any commit, while running the npm publish command only when a new tag has been committed to the master branch.

I could imagine that such an example might really improve the adoption of Azure CI as build solution for npm-based Javascript projects. Right now it is very difficult to find useful documentation for this individual problem.

Thanks in advance :)

copy from: microsoft/azure-pipelines-agent#1749

Have you tried trouble shooting?
Queuing a build generates an error. Unexpected value 'condition'.

Agent Version and Platform
Version of your agent? 2.102.0/2.100.1/...

OS of the machine running the agent? Windows

VSTS Type and Version
VSTS

If VisualStudio.com, what is your account name? https://fsmb.visualstudio.com

What's not working?
I am trying to set up a YAML build. We have a lot of builds and our existing definitions use task groups. I am moving the task group logic into a template so they can be shared across build definitions. The templates are stored in a separate repository. All this is working correctly.

There is one step that shouldn't execute when it is a PR build so I tried to do this.

• template: sharedstep.yml@templates
  parameters:
  value: true
  condition: and(succeeded(), eq(variables['Build.Reason'], 'PullRequest'))
  The build won't queue because of the error - Unexpected value 'condition'.

It would be nice if templates could be conditional like tasks are. Right now the workaround would be to either use a parameter or specify the condition inside the template but the template may not always know the correct condition to use.

Agent and Worker's Diagnostic Logs
It is not queuing so there is no log.

Related Repositories
Please ensure you are logging issues to the correct repository in order to get the best support.

Tasks Repository - contains all of the inbox tasks we ship with VSTS/TFS. If you are having issues with tasks in Build/Release jobs (e.g. unreasonable task failure) please log an issue here.
Hosted Agent Image Repository - contains the VM image used in the VSTS Hosted Agent Pool. If you are having Build/Release failures that seems like they are related to software installed on the Hosted Agent (e.g. the DotnetSDK is missing or the AzureSDK is not on the latest version) please log an issue here.
If you are hitting a generic issue about VSTS/TFS, please report it to the Developer Community

Description

I have a job in a template that depends on a preceding job as well as a template parameter:

parameters:
  runTests: true

# omitted for brevity

- job: "Test"
  dependsOn: "Build"
  condition: ${{and(parameters.runTests, succeeded())}}

I'm trying to use the succeeded() job status function within a template expression to ensure that the job only runs if "Build" succeeded and "runTests" is true. Unfortunately, it seems like this isn't allowed - the job "Test" fails with an error message stating that "succeeded" is invalid.

I've tried

and(${{parameters.runTests}}, succeeded())

as well.

Is there a way to evaluate parameters and job status functions in a single boolean expression?

By "cross-product matrix", I mean the scenario where you want to use the matrix strategy on a cross-product of multiple dimensions. For example, say you want to run tests on 2 OS (Linux, Windows) and 2 versions of Python (3.6, 3.7). Currently, you'd need to manually create a matrix with all the combinations:

strategy:
  matrix:
    Linux_Python36:
      VM_IMAGE: 'ubuntu-16.04'
      PYTHON_VERSION: '3.6'
    Linux_Python37:
      VM_IMAGE: 'ubuntu-16.04'
      PYTHON_VERSION: '3.7'
    Windows_Python36:
      VM_IMAGE: 'vs2017-win2016'
      PYTHON_VERSION: '3.6'
    Windows_Python37:
      VM_IMAGE: 'vs2017-win2016'
      PYTHON_VERSION: '3.7'

While this isn't too bad with a small number of dimensions and a small number of variables in each dimension, it can quickly become unmaintainable as the dimensions and variables grow. It would be nice if there was a way to express this more succinctly, something like:

strategy:
  cross-product:
    PYTHON_VERSION: [ '3.6', '3.7' ]
    VM_IMAGE: [ `ubuntu-16.04`, `vs2017-win2016` ]

logs from azure-pipelines show stdout output out of order; this seems like a serious bug.

I created a minimal example repo and pipeline so you can reproduce this issue:

see repo: https://github.com/timotheecour/timcitest (at revision f61677330cb33d781844bdb46f115165af6243ae)
see logs: https://dev.azure.com/timotheecour/timotheecour/_build/results?buildId=88

python start1 should appear before Cloning into 'csources', but it doesn't, see below:

2019-01-20T05:40:43.1789050Z ##[section]Starting: Run a multi-line script
2019-01-20T05:40:43.1792367Z ==============================================================================
2019-01-20T05:40:43.1792439Z Task         : Command Line
2019-01-20T05:40:43.1792486Z Description  : Run a command line script using cmd.exe on Windows and bash on macOS and Linux.
2019-01-20T05:40:43.1792579Z Version      : 2.146.1
2019-01-20T05:40:43.1792623Z Author       : Microsoft Corporation
2019-01-20T05:40:43.1792673Z Help         : [More Information](https://go.microsoft.com/fwlink/?LinkID=613735)
2019-01-20T05:40:43.1792765Z ==============================================================================
2019-01-20T05:40:44.9491923Z Generating script.
2019-01-20T05:40:44.9551007Z [command]/bin/bash --noprofile --norc /home/vsts/work/_temp/ef990781-b2bb-4666-9c12-235027eabc79.sh
2019-01-20T05:40:44.9716813Z Add other tasks to build, test, and deploy your project.
2019-01-20T05:40:44.9717682Z See https://aka.ms/yaml
2019-01-20T05:40:45.0116585Z Cloning into 'csources'...
2019-01-20T05:40:49.9792667Z python start1
2019-01-20T05:40:49.9793315Z in buildNimCsources
2019-01-20T05:40:49.9794349Z ('runCmd', 'git clone --depth 1 https://github.com/nim-lang/csources.git')
2019-01-20T05:40:49.9794599Z output:
2019-01-20T05:40:49.9794735Z 
2019-01-20T05:40:49.9794887Z after buildNimCsources
2019-01-20T05:40:50.0340286Z ##[section]Finishing: Run a multi-line script

I've reduced this from a more complex example where the out of order logs were making debugging very difficult.

EDIT

looks like it's due to stderr buffering (prob line buffering) become out of sync w stdout buffering (prob block)
workaround: unbuffer or python -u

Hello,

I'm trying to do something like this, which works for other variables:

• ${{ if ne(variables['Build.Repository.Provider'], 'GitHub') }}:

Basically, do some logic if this isn't a github build. However, Build.Repository.Provider doesn't seem to resolve at yaml "compile" time. If I check for 'github' in my build definition name (hacky) it works.

The build that’s trying to run microbuild is at https://dnceng.visualstudio.com/internal/_build/results?buildId=37125&_a=summary&view=logs
Notice the Install Microbuild Plugin step does run and it’s not supposed to.

I downloaded the binlog for the debug phase and the Build Repository Provider is indeed set as it should be.

Is this a bug?

I've put together a template for R that I think others would be interested in. I'll create a pull request for you to review.

As seen on stackoverflow.

I have the following pipeline that runs a container job using an image from a private azure registry:

resources:
  containers:
  - container: build_container
    image: gxg08regtest.azurecr.io/ci-build-image:66162
    endpoint: SandboxGXG08

pool:
  vmImage: 'ubuntu-16.04'

container: build_container

steps:
- script: ./build.sh

This seems close enough to this snippet but it gives me the following error:

Expected 'dockerregistry' service connection type for image registry referenced by build_container, but got azurerm for service connection SandboxGXG08.

Any idea?

While working with Ruby build/test, tests were failing due to line ending issues.

I added a step before 'checkout' in the script with the following:

git config --global core.autocrlf false

That solved the issue. I guess I'm not sure what I think might be changed. Any or all of the following?

1. Change git core.autocrlf default to false, currently it's true.
2. Change the docs to make it clear that steps can be added before checkout.
3. Add more options to checkout yaml to control git parameters.
4. Make it clear that one can set checkout to none and use whatever git command one wants.

Re 2, the following is shown:

steps: [ script | bash | powershell | checkout | task | stepTemplate ]

So it is doc'd...

Copied from: https://github.com/MicrosoftDocs/vsts-docs/issues/2327

It would be nice if you could dynamically set the display name of a task. This would allow the build logs to show information more relevant to the build.

For example you might have a build task that generates a version number for the build. It would be nice to be able to see the version number in the log summary by updating the display name. This would require the display name to be able to reference parameters, variables, env vars or something equivalent that preceding steps could set.

Based upon the link given above it is possible by using separate jobs. But I'd like to see this support added directly so we don't have to create separate jobs just to set the display name.

I am currently running into an issue where I am able to pull from one private GitHub source in the checkout stage but unable to pull another private GitHub source from a script later in the pipeline due to lack of access rights. Note: Both Repos are a part of the same organization. And I've already connected the Azure App to GitHub with access to all Repos

Within my checkout stage I've explicitly set persistCredentials: true This, as I understand it, should allow the following scripts in the pipeline to use the GitHub credentials that were used in the checkout for "Get Sources".

Here is an example of the script that fails:

- script: |
git clone --branch=username --single-branch https://github.com/username/myRepo.git $(Agent.BuildDirectory)/myRepo
displayName: 'clone myRepo' 

And the output:

Generating script.
[command]/bin/bash --noprofile --norc /Users/vsts/agent/2.140.2/work/_temp/cb2622cc-28e0-435a-bb98-154bdabf9641.sh
Cloning into '/Users/vsts/agent/2.140.2/work/1/myRepo'...
fatal: could not read Username for 'https://github.com': Device not configured
##[error]Bash exited with code '128'

Provide a way to select agent queues with matrix the same way as containers described here:
Container jobs - Multiple jobs

Use case - I have .NET Core library that needs to be built and tested on all platforms - Windows, Linux & Mac. Right now I can solve it with Job reuse with parameters but I think matrix would be more "natural" and less verbose way to do that

Moved from here: #1815

Hello,
I'd like to start an android emulator in my Pipelines yml. I followed the instructions on MSFT to create a bash script, but the build fails:

Invalid file path '/Users/vsts/agent/2.144.0/work/1/s/buildmachine'.

What did I do wrong?

Thanks,
Igor

The PowerShell step (backed by the PowerShell task) is limited to 2000 characters. This prevents a task from doing a reasonable amount of work since it is too big. There needs to be a way to run scripts larger than 2000. The possibilities that come to mind.

1. Use an external file. Unfortunately this doesn't work when you are using step templates at this time. So you cannot create a reusable template that contains PowerShell scripts larger than 2K.
2. Break up the script into separate steps. This is more doable except values calculated in one step would need to be recalculated in the next. Additionally the VSTS log shows all the steps in the output. So breaking up a logical step into child steps results in extra logs when run.
3. Switch to another shell that may not have the same limits. Undesirable since PS is the best choice for us.
4. Create multiple steps that write the script to a temp file, then load the temp file and run it. Seems like a hack and I'm not sure it is even doable.

I understand that the limitation is probably with the PS task. Nevertheless using large PS tasks in YAML seems more common than using it in a GUI build definition so I think something YAML-specific would need to be done to support this scenario.

I am trying to use the GitHub App authentication in multiple projects for the same organization.

I have created a project using the GitHub App (which is installed in GitHub giving access to all repos in the organization). This works fine as expected.
However, when I create another (a second) new project, I am prompted to use OAuth:

If I click on the other option Install our app from the GitHub Marketplace I am redirected to the GitHub App page:

Where I can confirm it is installed, and enabled for all repositories.

Expected behaviour
I would expect Azure Pipelines to give me a list of all my repositories when creating a new project.
According to the documentation:

The app will become Azure Pipelines’ default method of authentication to GitHub (instead of OAuth) when organization members create pipelines for their repositories. This is recommended so that pipelines run as “Azure Pipelines” instead of a user’s GitHub identity which may lose access to the repository.

I would like to run a specific template if an array of parameters exist or not:

#template.yml

• ${{ if eq(parameters.deploy_extra_vars, '') }}:

  • template: deploy.yml

• ${{ if ne(parameters.deploy_extra_vars, '') }}:

  • template: deploy2.yml

#vsts-ci.yml
parameters:
deploy_extra_vars: |
test: "testing"
test1: "testing1"

Now if in my vsts-ci.yml file I do not pass the 'deploy_extra_vars' parameter I receive an error. If it exists run this template, if it does not run this template.

See microsoft/azure-pipelines-agent#1733
for more info.

I have the following parameter in a yaml build and it's set to 'yes' in VSTS variables tab.

parameters:
  buildDebugPackage: '$(BuildDebugPackage)'

I would like to do the following

- ${{ if eq(parameters.buildDebugPackage, 'yes') }}:
  - ${{ parameters.preBuildDebugProjects }}
  - task: DotNetCoreCLI@2
    displayName: Build Debug Projects
    inputs:
      projects: ${{ parameters.projects }}
      arguments: '--no-restore -c Debug /p:Version=$(ASSEMBLY_VERSION);FileVersion=$(FILE_VERSION)'
  - ${{ parameters.postBuildDebugProjects }}

But this does not work because the variable BuildDebugPackage is not expanded to the value 'yes'.

I have to instead do the following

- powershell: |
    Write-Host "##vso[task.setvariable variable=BuildDebugPackage]$($env:BuildDebugPackage)"
  displayName: Setup Environment Variables
  env:
    BuildDebugPackage: ${{ parameters.buildDebugPackage }}
  ignoreLASTEXITCODE: false
  errorActionPreference: Stop
  failOnStderr: true

- ${{ parameters.preBuildDebugProjects }}
- task: DotNetCoreCLI@2
  displayName: Build Debug Projects
  condition: and(succeeded(), eq(variables[BuildDebugPackage], 'yes'))
  inputs:
    projects: ${{ parameters.projects }}
    arguments: '--no-restore -c Debug /p:Version=$(ASSEMBLY_VERSION);FileVersion=$(FILE_VERSION)'
- ${{ parameters.postBuildDebugProjects }}

This is far inferior as I cannot even condition the pre and post steps. This also shows up on the UI as a skipped step instead of not showing up at all. I also have to convert the parameter to a variable in PowerShell as well. Please fix this bug or give a workaround. Thanks.

Even though docs state that the checkout function allows cloning a secondary repo, when I try running a simple pipeline to push to a secondary repo which is within the same user account, I get this error Checkout of repository 'destinationRepo' is not supported. Only 'self' and 'none' are supported.

Here's the code I'm trying to test with:

trigger:
- master

pool:
  vmImage: 'Ubuntu-16.04'

resources:         
  repositories:
  - repository: destinationRepo
    type: github
    name: samiyaakhtar/solid-doodle
    endpoint: solid-doodle-push-auth

steps:
- checkout: destinationRepo
  persistCredentials: true
  clean: true

- powershell: |
   git --version
   git checkout master
   git config user.email "[email protected]"
   git config user.name "AzureDevOps Bot"
   Write-Host "git commit with message"
   git commit  --allow-empty -a -m "Test Commit from Azure DevOps"
   Write-Host "git push"
   git push

copy from: microsoft/azure-pipelines-agent#1809

When I try to build my source using the yaml below;

pool:
vmImage: 'Ubuntu 16.04'

variables:
imageName: 'xxxredactedxxx'
projectfolder: 'xxxredactedxxx'

steps:

• script: docker build -f $(projectfolder)/Dockerfile -t $(imageName):$(build.buildId) -t $(imageName):latest .
  displayName: 'docker build'

• task: AmazonWebServices.aws-vsts-tools.ECRPushImage.ECRPushImage@1
  displayName: 'Push Image: '
  inputs:
  awsCredentials: 'AWS'
  regionName: 'eu-west-1'
  sourceImageName: '$(imageName)'
  repositoryName: '$(imageName)'
  autoCreateRepository: true
  I get the following error;

Job phase1: Step input awsCredentials references service connection AWS which could not be found. The service connection does not exist or has not been authorized for use. For authorization details, refer to https://aka.ms/yamlauthz

I confirm I have a service connection named AWS.

I tried reading the link in the error message. However, the article doesn't show any directions on how to authorize the build to use the service conneciton.

This npm script tasks in node templates fail on Windows agents:

- script: |
    npm install
    npm run build

https://github.com/Microsoft/azure-pipelines-yaml/blob/44d3c750751e8f06177b6beaffee0d8fa3118ae2/templates/node.js.yml#L15-L17

To repro, change the vmImage to vs2017-win2016 and trigger a build.

I believe this happens because:

• npm is npm.cmd on Windows
• The script task combines npm install and npm test into a single .cmd file and executes it
• When an outer batch file calls an inner batch file (without using call syntax), when the inner batch file returns, the outer batch file exits early
• So, the generated .cmd file exits after the first npm command

You can repo this on any Windows machine by executing a batch file containing:

npm install
npm test

To execute both commands, the call syntax must be used:

call npm install
call npm test

I believe the best fix is to use the Npm task instead of the script task. Here's an issue and pull request for the same issue in the pipelines-javascript sample repo:

MicrosoftDocs/pipelines-javascript#10
MicrosoftDocs/pipelines-javascript#3

Here are all the impacted templates in this repo:

https://github.com/Microsoft/azure-pipelines-yaml/search?q=npm&unscoped_q=npm

YAML supports PowerShell and a couple of other scripting languages. It would be nice if it could support TypeScript as well. There appears to be a shift in preference for all the DevOps tasks to use TypeScript instead of PowerShell. So it would be nice if we could move our TypeScript code into YAML and have it called directly in YAML instead of PowerShell.

For this to work we'd need better integration than we currently have in PS. At a minimum external scripts would need to be supported. Additionally we would need the Task SDK library available so we can call the various DevOps functions that are available in the SDK.

Exactly how this would work I'm not sure but I could envision a call to the TS compiler followed by it being run sort of like how VS Code can run TS now.

Is it possible to use a variable or parameters in agent pool demands? Something like this:

variables:
  projectName:my-vsts-project
  agentName: '$(projectName)-$(Build.BuildId)'

jobs:
- job: JavaBuild
  dependsOn: LaunchAgent
  pool:
    name: 'mgmt-aks-sandbox'
    demands:
    - agent.name -equals $(agentName)

It doesn't seem to work, as I end up getting like this:

##[Error 1]
No agent found in pool mgmt-aks-sandbox which satisfies the specified demands:
     agent.name -equals $(projectName)-$(Build.BuildId)
     java
     Agent.Version -gtVersion 2.140.2 

I've also tried with parameters and template expressions, and those don't work either:

##[Error 1]
No agent found in pool mgmt-aks-sandbox which satisfies the specified demands:
     agent.name -equals ${{ parameters.agentName }}
     java
     Agent.Version -gtVersion 2.140.2 

It looks like this guy managed to do it via the UI (see comments at the end), but is there a way to do it in YAML?
https://www.noelbundick.com/posts/serverless-vsts-build-agents-with-azure-container-instances/

First of all, I worked with visual designer last year, and it was cool.

now with the yaml is even much more cooler & flexible. I love that.

But I have a big question, that I wasn't able to answer myself looking at the good docs here.

so, I want to have this situation:

• I want a CI only for the PRs towards develop branch, which should only build & run tests
• I want a CD that runs only when develop branch is updated, which should build & run tests & publish artifacts to a DEV Release
• I want a CD that runs only when master (releases) branch is updated, which should build & run tests & publish artifacts to a PROD release

and now my question is: how do I define different pipelines for different branches? I wasn't able to have multiple yaml, which would be ideal, or put some 'conditions'

What I'm getting wrong here? It's a missing thing?

Please add pull request triggers to yaml (currently not supported).

To emulate task groups in regular build definitions we have created YAML steps in a build repository that all our other builds reference using the reusable step syntax given in the documentation. This works correctly for us.

However some of the logic either cannot be done directly in YAML or doesn't work well because of issues in the current implementation. So we figured the best solution was to pull that logic into a separate script (.ps1) that sits next to the .yaml file. For non-reusable steps this seems to work OK but for steps that reused in other repos we cannot figure out the path that works. We have even gone so far as to dump the entire build directory to see where the yaml files go and we cannot see that the reusable steps (in other repos) are ever copied to the build directory such that a path would work.

So, what is the correct path to use when you want to call a .ps1 script from a YAML file that is called from other YAML files in other repos?

- powershell: ${{ format('.\steps\my-step.ps1 {0}', parameters.someArg) }}

This appears to work if you do this in a YAML file that is in your repo you're building. It does not work if you try to do this from a YAML file that is in a reusable step.

- powershell: ${{ format('$(Build.Repository.LocalPath)\steps\my-step.ps1 {0}', parameters.someArg) }}

Tried this approach as we saw something like this as part of one of the MS repos in GitHub (e.g. Azure build task or something).

Copy from microsoft/azure-pipelines-agent#1727

Have you tried trouble shooting?
Couldn't queue the build

Agent Version and Platform
Version of your agent?
2.136.1

OS of the machine running the agent?
CentOS 7 x64 7.5

VSTS Type and Version
VSTS

What's not working?
According to templateexpressions there is a possibility to use task condition for the templates. I've tried to use next value for one of the task input ${{ and(not(parameters.force), not(parameters.override)) }} and it wasn't calculated properly. Limiting value to just one function ${{ not(parameters.force) }} throws the same error. I want to have task enabled/disabled depending on the value of two parameters. Is it possible?

Error in VSTS while trying to queue a build:
build.package.yml (Line: 13, Col: 14): Unexpected type: 'System.Boolean',build.package.yml (Line: 13, Col: 14): Expected a Boolean value. Actual value: ''

.vsts-ci.yml

name: $(SourceBranchName)_$(Date:yyyyMMdd)$(Rev:.rr)
queue: PrivatePool
steps:

• template: build.package.yml
  parameters:
  force: true

build.package.yml

parameters:
force: false
override: false

steps:

• checkout: self
  clean: true

• task: DownloadBuildArtifacts@0
  displayName: Download Build Artifacts
  condition: and(succeeded(), in(variables['build.reason'], 'Manual', 'Schedule'))
  continueOnError: true
  enabled: ${{ and(not(parameters.force), not(parameters.override)) }}
  inputs:
  buildType: specific
  downloadType: specific

In my PowerShell script I'm trying to generate a copyright message. I put © into a string and it gets converted to a lowercase c when the script is run (in the log and in the actual script).

# This will be a c
$copyright = "Copyright ©"

I also tried using the raw Unicode equivalent.

# This will also be a c
$copyright = "Copyright " + [char]0x00A9

This works in PowerShell directly but not when put into a PowerShell task in YAML.

When using the full syntax for jobs the timeoutInMinutes can be set at the same level as the job. From the docs:

jobs:
- job: string
  timeoutInMinutes: number
  cancelTimeoutInMinutes: number
  strategy:
    maxParallel: number
    # note: `parallel` and `matrix` are mutually exclusive
    # you may specify one or the other; including both is an error
    parallel: number
    matrix: { string: { string: string } }
  pool:
    name: string
    demands: string | [ string ]

  container: string

  steps:
    - script: echo Hello world

However if I set it when using the single-job syntax the build pipeline fails with error "Unexpected value 'timeoutInMinutes'"

If I move the timeoutInMinutes into the pool element it works ok.

If this is by design then it might be worth updating the docs to indicate this.

Thanks :)

The YAML parsing reports errors when trying to queue a YAML build but often the actual error is well beyond the truncated error message that is shown. It would be nice if the YAML parser would provide a more focused error message around the actual code (ideally with lines if applicable) that is causing the error instead of just showing the first part of the script.

Example 1 - Script is longer than 2000 characters. The only way to figure out how much you have to truncate is to load the script into an editor that counts characters. What makes this difficult at best is the fact that it is unclear what counts to the parser. I noticed that template expressions cause the entire block to be put inside a format message where the expressions are replaced with ordinal positions. So is this part of the char limit or the original template expression? What about the spaces before the start of the line when doing something like a multiline PowerShell script where you have to put spaces to get the indentation proper? Rather than memorizing, playing around with the rules it would be nice if it would just show the line and an indicator where it ran out of room.

Example 2 - Body has ${ elements such that parser says a $ is invalid. This is a complex PowerShell script mixing template expressions, PowerShell variables, a hashtable, regular expression string with nested quotes and $. The only message I get is that the $ is invalid but there is no easy way to figure out where this is actually occurring. Note that line/column information probably isn't sufficient here since the code is converted to a format call. It would be unclear whether this is the original code or the modified code.