dotCover not generating coverage for all assemblies - msbuild

I've set up dotCover to run using an .xml
<?xml version="1.0" encoding="utf-8"?>
<CoverageParams>
<TargetExecutable>
c:\dotcover\xunit\xunit.console.exe
</TargetExecutable>
<TargetArguments>
"INWK.Configuration.UnitTests.dll"
</TargetArguments>
<TargetWorkingDir>
..\bin\x64\Debug\
</TargetWorkingDir>
<TempDir>
<!-- Directory for auxiliary files. Set to the system temp by default. -->
</TempDir>
<Output>
dotCover-xunit.dcvr
</Output>
<InheritConsole>
<!-- [True|False] Lets the application being analyzed to inherit dotCover console. True by default. -->
</InheritConsole>
</CoverageParams>
You can see (Service, Shared, UnitTests assemblies correctly included in the test coverage report (Shared, Service and UnitTest assemblies)
However, when running the same on the build server *Service and *Shared are missing.
After replacing Service.dll and Shared.dll and their "pdb's" from local copy to build server and running dotCover on build server again it works correctly.
This leads me to believe that build server runner does something different than msbuild.exe from VS when running build locally.
I found very similar issue description here: https://stackoverflow.com/questions/25855131/dotcover-and-xunit-not-gathering-coverage-statistics-in-some-environments, but not sure how to remedy this in my build server configuration.
Trace log output (one drive)
https://1drv.ms/t/s!AtxuuqGHIqXwgTVqQJ_Y_-rGE8W9?e=HrZgj7

Found the solution:
in my dotcover config xml I had to add: -noshadow switch, like so:
<CoverageParams>
<TargetExecutable>
c:\dotcover\xunit\xunit.console.exe
</TargetExecutable>
<TargetArguments>
"INWK.OrderIndexing.UnitTests.dll" -noshadow
</TargetArguments>
<TargetWorkingDir>
..\bin\x64\Release\
</TargetWorkingDir>
...
Now all assemblies (except the ones I do want to filter) are showing up

Related

Launch IIS Express to run ASP.NET Core Apps

I wish to run my ASP.NET Core App by launching it from IIS Express using command line.
I stumbled across this article which says
So in fact Visual Studio silently adds the two environment variables
when launching IIS Express, so that ASP.NET Core related bits can be
injected.
LAUNCHER_ARGS: -debug -p “C:\Program Files\dotnet\dotnet.exe” -a “exec
\”C:\Users\lextm\documents\visual studio
2017\Projects\WebApplication2\WebApplication2\bin\Debug\netcoreapp1.0\WebApplication2.dll\””
-pidFile “C:\Users\lextm\AppData\Local\Temp\2\tmpFD6D.tmp” -wd “C:\Users\lextm\documents\visual studio
2017\Projects\WebApplication2\WebApplication2”
The tmp file in -pidFile “C:\Users\lextm\AppData\Local\Temp\2\tmpFD6D.tmp” can always change. How do I add LAUNCHER_ARGS as environment variable which will make it work even if the tmp file changes?
Let me know if there is any easier way to launch IIS Express to run ASP.NET Core Apps with command line or powershell scripts.
You are looking for [System.IO.Path]::GetTempFileName() method. It creates empty temp file on file system and returns its unique name.
I'm currently using the following PowerShell script to run my .NET Core 2.0 App:
$env:LAUNCHER_ARGS = "-p ""<path to dotnet.exe>"" -a ""exec \""<path to webapp main dll>\"""" -pidFile $([System.IO.Path]::GetTempFileName()) -wd ""<path to webapp root folder>"" -pr <project name>"
$env:LAUNCHER_PATH = "<path to VSIISExeLauncher.exe>"
& "<path to iisexpress.exe>" /config:"<path to applicationhost.config>" /site:"<webapp name>"
Placeholders (text within angle brackets) have to be filled with the corresponding values. You can find them out by running your project from Visual Studio and inspecting environment variables of iisexpress.exe process using Process Explorer as shown above in the link you provided.
In .NET Core 3 the solution to this problem has changed. Follow these steps.
1) The environment variables should now be:
LAUNCHER_ARGS=exec "C:\YourWebApiProject\bin\Debug\netcoreapp3.1\YourWebApiProject.dll"
LAUNCHER_PATH=C:\Program Files\dotnet\dotnet.exe
Change the first path to your dll path and ensure the .NET version in the path is correct. Note that there is no longer a need to create a temp file.
2) Ensure that both modules AspNetCoreModule and AspNetCoreModuleV2 are registered in the file .vs\{your solution name}\config\applicationhost.config as follows:
Under <system.webServer> <globalModules> add:
<add name="AspNetCoreModule" image="%IIS_BIN%\aspnetcore.dll" />
<add name="AspNetCoreModuleV2" image="%IIS_BIN%\Asp.Net Core Module\V2\aspnetcorev2.dll" />
Under <sectionGroup name="system.webServer"> add:
<section name="aspNetCore" overrideModeDefault="Allow" />
Under <location path="" overrideMode="Allow"> <system.webServer> <modules> add:
<add name="AspNetCoreModule" lockItem="true" />
<add name="AspNetCoreModuleV2" lockItem="true" />
It's also a good idea to make this change to the templates for this file which are located at %PROGRAMFILES%\IIS Express\config\templates\PersonalWebServer\applicationhost.config and %PROGRAMFILES(x86)%\IIS Express\config\templates\PersonalWebServer\applicationhost.config so that new VS solutions you create automatically get these changes to their configs. (Credit to this post)
3) Be mindful of whether you're using 32 or 64 bit IIS Express. (If you're on a 64 bit machine then 32 bit IIS Express = C:\Program Files (x86)\IIS Express, 64 bit = C:\Program Files\IIS Express) In my case, 32 bit had worked fine previously but after migrating to .NET Core 3 I had to use 64 bit or else the above modules wouldn't load.
I needed to run multiple .Net Core API endpoints at a time easily without popping open Visual Studio for each and every one of them. I ended up using answers here to build the following:
iisaspnet.bat:
#echo off
:: Args are like:
:: MobileApi
:: C:\Users\chris\Dropbox\Code\2017\VbaMeasureHcp\src\HCPMobileApi
:: MobileApi\bin\Debug\netcoreapp2.2\MobileApi.dll
:: .vs\MobileApp\config\applicationhost.config
setlocal
set PATH=%PATH%;C:\Program Files\IIS Express
set LAUNCHER_ARGS=exec %2\%3
set LAUNCHER_PATH=C:\Program Files\dotnet\dotnet.exe
iisexpress /site:%1 /config:"%2\%4"
:: Comment out line below to check for errors
exit
The first arg is the name of the Project - the name in Visual Studio (in some scenarios people go rogue and name their Project file one thing and the Project itself another thing - you want the Project name, not the file name, in this scenario).
The second arg is the root folder for third and fourth args.
The third arg is where to find the compiled Project DLL
The fourth arg is where to find the applicationhost.config that explains how to launch the site. As you'll see below, this is generally found in your .vs folder, but, where this exists can get a little crazy depending on how creative people get with organizing their Solution and Project folders. Generally the .vs folder is going to sit in the same folder as the .sln file, so it may be far from the Project folder/files.
This will be less helpful, but here's the batch file that kicks off the IIS Express windows, so you can see iisaspnet.bat in use:
start iisaspnet.bat MobileApi C:\Users\chris\Dropbox\Code\2017\VbaMeasureHcp\src\HCPMobileApi MobileApi\bin\Debug\netcoreapp2.2\MobileApi.dll .vs\MobileApp\config\applicationhost.config
start iisaspnet.bat HCP.MasterDataApi C:\Users\chris\Dropbox\Code\2017\VbaMeasureHcp\src MasterData\Api\HCP.MasterDataApi\bin\Debug\netcoreapp2.2\HCP.MasterDataApi.dll Solutions\.vs\MasterData\config\applicationhost.config
start iisaspnet.bat HCP.SecurityApi C:\Users\chris\Dropbox\Code\2017\VbaMeasureHcp\src Security\Api\HCP.SecurityApi\bin\Debug\netcoreapp2.2\HCP.SecurityApi.dll Solutions\.vs\Security\config\applicationhost.config
start iisaspnet.bat HCP.BillingApi C:\Users\chris\Dropbox\Code\2017\VbaMeasureHcp\src Billing\Api\HCP.BillingApi\bin\Debug\netcoreapp2.2\HCP.BillingApi.dll Solutions\.vs\Billing\config\applicationhost.config
start iisaspnet.bat HCP.ClientApi C:\Users\chris\Dropbox\Code\2017\VbaMeasureHcp\src Client\Api\HCP.ClientApi\bin\Debug\netcoreapp2.2\HCP.ClientApi.dll Solutions\.vs\Client\config\applicationhost.config
start iisaspnet.bat HCP.EmployeeApi C:\Users\chris\Dropbox\Code\2017\VbaMeasureHcp\src Employee\Api\HCP.EmployeeApi\bin\Debug\netcoreapp2.2\HCP.EmployeeApi.dll Solutions\.vs\Employee\config\applicationhost.config
The way the parameters work could likely be far simpler if the way the Projects, Solutions, etc were stored and named was more consistent, but, this is an existing set of Solutions I had no control over with scattered naming etc, and, the chaos above may be more helpful anyway for understanding how to call these commands.
The result is that running one command kicks off 6 IIS Express command windows for me, requests get logged to each of their windows, and I just type Q in each window to kill them.

Read dotCover total coverage output from GitLab CI

I configured dotCover to be run from our GitLab CI server.
It correctly runs the tests, produces the required output and the CI is configured to store the coverage HTML output in the GitLab artifacts. This works flawlessly.
What I'm trying to do is to read the total coverage output from the dotCover.exe console runner and parse it in gitlab CI. I read the dotCover documentation but I did not find a method to output a line containing the coverage to console. Gitlab CI can only read coverage values from the sdout of the ci job, matching it with a custom regex.
This is my config.xml:
<?xml version="1.0" encoding="utf-8"?>
<AnalyseParams>
<TargetExecutable>C:\NUnit\nunit3-console.exe</TargetExecutable>
<TargetArguments>--agents=1 MyDll.Spec.dll MyDll2.Spec.dll</TargetArguments>
<Output>cover/AppCoverageReport.html</Output>
<ReportType>html</ReportType>
<Scope>
<ScopeEntry>MyApp\bin\x86\Release\net461\MyApp.*.dll</ScopeEntry>
<ScopeEntry>MyApp\bin\x86\Release\net461\*.exe</ScopeEntry>
</Scope>
<Filters>
<ExcludeFilters>
<FilterEntry>
<ModuleMask>*.Spec</ModuleMask>
</FilterEntry>
</ExcludeFilters>
</Filters>
</AnalyseParams>
and I run it with .gitlab-ci.yml:
C:\dotCover\dotCover.exe analyse config.xml /TargetWorkingDir=.
Is there a way to view this value in GitLab CI? Am I missing something obvious?
Thank you
I ended up reading the HTML output file of dotCover and parsing the output.
Luckily the total coverage is at an easily parse-able portion of the output file. The coverage regex is
'/= \[\["Total",\d+\.?\d+/'
This is my final .gitlab-ci.yml file (for Windows runner, dotCover is windows-only):
my_job:
# your job configuration
# ...
scripts:
# build the solution here, ...
- dotCover.exe analyse dotCover.xml /TargetWorkingDir=.
- type cover\AppCoverageReport.html
coverage: '/= \[\["Total",\d+\.?\d+/'
Not a very long-term solution but it works for now, at least until I update dotCover.

commandline tool Glassfish appclient doesn't see deployed EJB

I'm working my way through a book about "Java EE 7 for Glassfish", with the server installed on Fedora Linux.
I have a simple stateless session bean SimpleSessionBean deployed on the server and I am trying to approach that SimpleSessionBean via SessionBeanClient and the Glassfish command line tool appclient, running a client jar. Everything from the book, so it should work. The client however can't find SimpleSessionBean. Apparently a class path issue. In the server logs nothing happened.
I can't find any pointers how Glassfish should be properly installed. Everything works within the server. I can approach installed war files from facelets running in a browser.
It is probably a matter of setting $PATH right or something or some other environment variable. Any pointers to relevant literature?
Thanks in advance for any suggestions!
UPDATE1: error message
From the bash terminal window where I run appclient:
[fedora#localhost bin]$ ./appclient -client /home/fedora/Downloads/6886EN_04_Code/ch04_src/simplesessionbeanclient/target/simplesessionbeanclient.jar
Jul 06, 2017 12:52:57 PM org.glassfish.apf.impl.DefaultErrorHandler error
SEVERE: Class [ Lnet/ensode/glassfishbook/SimpleSession; ] not found.
Error while loading [ class net.ensode.glassfishbook.SessionBeanClient ]
Exception in thread "main" java.lang.NoClassDefFoundError: net/ensode/glassfishbook/SimpleSession
at net.ensode.glassfishbook.SessionBeanClient.invokeSessionBeanMethods(SessionBeanClient.java:12)
at net.ensode.glassfishbook.SessionBeanClient.main(SessionBeanClient.java:19)
Caused by: java.lang.ClassNotFoundException: net.ensode.glassfishbook.SimpleSession
at java.net.URLClassLoader.findClass(URLClassLoader.java:381)
at org.glassfish.appclient.client.acc.ACCClassLoader.findClass(ACCClassLoader.java:237)
at java.lang.ClassLoader.loadClass(ClassLoader.java:424)
at java.lang.ClassLoader.loadClass(ClassLoader.java:357)
UPDATE2: From the Glassfish book:
We are using Maven to build our code. For this example, we used
the Maven Assembly plugin (http://maven.apache.org/plugins/maven-assembly-plugin/) to build a client JAR file
that includes all dependencies; this frees us from having to specify
all the dependent JAR files in the -classpath command-line option
of the appclient utility. To build this JAR file, simply invoke mvn
assembly:assembly from the command line.
SOLUTION: the missing link was producing a client jar with additional jar's "on board" so to speak. Proceed as follows (at least in Eclipse): select pom.xml > right-click > Run As > Maven build... > enter in Goals field: assembly:assembly> Apply/Run.
The result will be that you will find TWO jars under your target folder: xxxclient.jar and xxxclient-jar-with-dependencies.jar.
From the command line in bash execute from the folder with latter jar:
/path_to/appclient -client xxxclient-jar-with-dependencies.jar
After a very long wait (on my $200 mini Linux box) the HelloWorld-ish server EJB gets finally properly called.
Your assumption is right.
You are missing net.ensode.glassfishbook.SimpleSession in your classpath.
From an older book online:
...executed through the appclient utility. This utility can be found at
[glassfish installation directory]/glassfish/bin/. Assuming this path
is in the PATH environment variable, and assuming we placed our client
code in a JAR file called simplesessionbeanclient.jar, we would
execute the above client code by typing the following command in the
command line:
appclient -client simplesessionbeanclient.jar
It seems that you've started from
.../bin/./appclient -client
/home/fedora/Downloads/6886EN_04_Code/ch04_src/simplesessionbeanclient/target/simplesessionbeanclient.jar
You need SimpleSession.class in your CLASSPATH (or in a jar in that classpath).
Usually java checks the current directory first (which is your bin folder). If the class is not found (its not, since its in your simplesessionbeanclient folder), it searches for that class in the classpath (where you did not add the simplesessionbeanclient folder).
Try
appclient -client simplesessionbeanclient.jar
from the folder where simplesessionbeanclient.jar is located.
If you don't want to add the appclient folder to your path start with
/your/path/to/appclient -client simplesessionbeanclient.jar
(again from the folder where simplesessionbeanclient.jar is located)
Update:
If you still get a ClassNotFoundException have a look if it is missing in your jar file (jars are Zip-File, you could use your Zip-Tools):
jar tf simplesessionbeanclient.jar
if there is a SimpleSession.class
I did the following to fix my problem:
Use appclient -classpath (instead of appclient -client)
Use the regular project JARs (instead of the one generated by mvn assembly:assembly)
Deploy the EJB to Glassfish (simplesessionbean.jar)
The example code from a more recent book "Java EE 8 Application Development" by David R. Heffelfinger (the same author of "Java EE 7 for Glassfish") is almost exactly the same (the only minor difference is classes are packaged in "net.ensode.javaee8book" instead of "net.ensode.glassfishbook").
When running appclient.bat -client simplesessionbeanclient-jar-with-dependencies.jar I kept getting:
java.lang.ClassNotFoundException: <mainclass>
errors. This was because the POM was assembling a manifest with <mainClass> value of "net.ensode.glassfishbook.SessionBeanClient" (instead of "net.ensode.javaee8book.SessionBeanClient"). So I decided to avoid using the -client option for appclient.bat and switched to -classpath which allowed me to specify the main class on the command line (which is easier than updating the POM or refactoring the packages to suit the manifest).
But then when running the appclient command:
PS C:\home\programs\java_ee_sdk-8u1\glassfish5\glassfish\bin> .\appclient.bat -classpath "C:\home\code\Java-EE-8-Application-Development-Code-Samples-master\ch04_src\simplesessionbean\target\simplesessionbean.jar;C:\home\code\Java-EE-8-Application-Development-Code-Samples-master\ch04_src\simplesessionbeanclient\target\simplesessionbeanclient.jar" net.ensode.javaee8book.SessionBeanClient
I kept getting:
Root exception is javax.naming.NameNotFoundException: net.ensode.javaee8book.SimpleSession#net.ensode.javaee8book.SimpleSession not found]]
errors. This was solved by deploying the EJB (simplesessionbean.jar) to Glassfish via the Admin Console (this missing step was not mentioned in the book). Running the appclient.bat command then worked.
Screenshot of appclient.bat (takes about 15 seconds to load):
Screenshot of EJB deployment:
Alternatively
You can manually compile the client to include all the dependencies by copying SimpleSession.java and SimpleSessionBean.java from the "simplesessionbean" project to the "simplesessionbeanclient" project (remember to refactor the package statements). This will generate simplesessionbeanclient.jar with the EJBs included (Nb: you still have to deploy the EJBs to the GlassFish server). Also make sure that the <mainClass> element in the POM points to the correct package.
You can now use the -client option:

SonarQube - integrationTest.exec - sonarRunner (Gradle) or "sonar-runner" command - showing 0.0% covereage

I'm successfully generating 2 .exec files by Jacoco within "build/jacoco" folder after running a Gradle based build and integration tests.
Gradle command:
"gradle clean build integrationTest"
Once done, it generates the following .exec files under build/jacoco folder.
test.exec
integrationTest.exec
Following is my sonar-project.properties file. When, I run "sonar-runner" from Linux prompt it completes but on SonarQube dashboard for this project, I see Unit test says some 34.5% but integration tests says 0.0%. Both .exec files have valid size. I also did "cat" on the .exec files and piped the output to "strings" command in Linux and saw that integrationTest.exec did hit the Tests functions - I have only 1 .java file.
When I run "gradle clean build integrationTest sonarRunner -Dxxx.xxx=yyy -Dyyy.xx=zzz" i.e. by passing all the sonar variable as mentioned in the sonar-project.properties file using -D option, it works but same result on SonarQube project's dashboard. Project's sonar dashboard has both widgets configured for Unit / Integration Tests and I'm including IT tests for showing Overall coverage. Overall coverage is showing 34.5% (which is Unit test % value). Sonar does see test.exec, integrationTest.exec and also auto generates overall-xxx.exec file as well during this operation.
NOTE: I'm no where - while starting tomcat on a separate putty / linux console -OR within Gradle build script, providing any value or setting JAVA Agent for Jacoco. I'm getting integrationTest.exec file and test.exec file already so not sure if JVM needs to be stopped once IT tests are complete running. I don't think I need those as i have valid file size for .exec files.
My ?:
- Why sonar is not getting IT coverage on the dashboard even though I'm setting / passing the following variable correctly:
sonar.jacoco.itReportPath=build/jacoco/integrationTest.exec
-bash-3.2$ cat sonar-project.properties
# Root project information
sonar.projectKey=com:company:product:ProjectA
sonar.projectName=ProjectA
sonar.projectVersion=1.0
# optional description
sonar.projectDescription=ProjectA Service
#Tells SonarQube that the code coverage tool by unit tests is JaCoCo
sonar.java.coveragePlugin=jacoco
#Tells SonarQube to reuse existing reports for unit tests execution and coverage reports
sonar.dynamicAnalysis=reuseReports
# Some properties that will be inherited by the modules
sonar.sources=src/java,test/java,src/java-test
# Sonar Unit Test Report path
sonar.jacoco.reportPath=build/jacoco/test.exec
# Sonar Integration Test Report Path
sonar.jacoco.itReportPath=build/jacoco/integrationTest.exec
sonar.junit.reportsPath=build/UT/results
# Sonar Binaries
sonar.binaries=build/classes/main
Narrowing down the cause: I think it's due to the .exec file for Integration test. To proove it: I passed UT exex file to both reportsPaths in Sonar variables i.e. the following and SonarQube picked both UT/IT test coverage. This prooves that if .exec file for IT tests is good (which I think it's But I need to double check) then Sonar will pick the .exec file and show a valid coverage % instead of 0.0%. Note: the following is just to proove if Sonar is picking the values or not. itReportPath variable should use the .exe file which is generated by Integration tests by Jacoco.
sonar.jacoco.reportPath=build/jacoco/test.exec
# Sonar Integration Test Report Path
#sonar.jacoco.itReportPath=build/jacoco/testintegrationTest.exec
sonar.jacoco.itReportPath=build/jacoco/test.exec
OK Found the issue. I was running integrationTest task in Gradle and was NOT attaching the jacocoagent.jar (as per Jacoco documentation) to the target JVM (Tomcat's instance) scope. Once I did that, I removed jacoco { ... } section from integrationTest task in Gradle (build.gradle or GRADLE_HOME/init.d/some.common.gradle file as this attach jacoco agent to the Java JVM in which Gradle is running). Now, once jacocoagent.jar was attached to Tomcat's JVM (as per the line below which I added in Tomcat's startup.sh script and added the variable to the command which starts Tomcat), then I ran Gradle (integrationTest) task for running IT tests.
PROJ_EXTRA_JVM_OPTS=-javaagent:tomcat/jacocoagent.jar=destfile=build/jacoco/IT/jacocoIT.exec,append=false
Then while Gradle was in progress, tests ran and I got a file (jacocoIT.exec at the given location) with some file size BUT this is not yet the final one. I had to stop the Tomcat session/JVM instance by running Tomcat's stop.sh script. Once Tomcat was stopped, I saw jacocoIT.exec file size increased significantly and this was the valid final jacocoIT.exec file (which I needed for sonarRunner Gradle task OR sonar-runner exectuable to pick and successfully push IT code coverage data to project's sonar dashboard). Once done, I got both UT + IT and it's combined code coverage.
sonar.jacoco.reportPath=build/jacoco/UT/jacocoUT.exec
sonar.jacoco.itReportPath=build/jacoco/IT/jacocoIT.exec

How can I remotely start/stop a service using MSBuild?

I'd like to remotely start or stop a windows service on another machine using MSBuild. To accomplish this, I wrote this script:
<Project xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
<Import Project="C:\Program Files (x86)\MSBuild\MSBuildCommunityTasks\MSBuild.Community.Tasks.Targets"/>
<Target Name="MyTarget">
<ServiceController MachineName="Box2" ServiceName="MyService" Action="Stop" />
</Target>
</Project>
When I run that on a machine that can see Box2, I get this:
Project
"C:\Scripts\Test.xml" on node 1 (default
targets).
C:\Scripts\Test.xml(4,5): error : Couldn't
find the 'MyService'
service on 'Box2' Done Building
Project
"C:\Scripts\Test.xml" (default targets) --
FAILED.
I know that I have the service name correct (I copied and pasted it from the actual service list), and I'm pretty sure that it can see Box2 because if I change it to a machine name that doesn't exist (e.g. Box2asdf), it takes about 10 seconds to come back (with the exact same error, mind you), as opposed to the nearly immediate response that I get when I provide the correct machine name.
How might I debug this issue?
You might try this instead...
You can use the command line program sc and execute that...
ie
SC \ServerName stop ServiceName
http://support.microsoft.com/kb/166819
For more information on how to execute a command from msbuild check this out..
execute a command with parameters using msbuild
The community tasks should work. Just use Sc query to check that the service does work. as for using msbuild its still using msbuild if you wrap sc in an exec?
At least you dont have a dependency on a third party dll in your build process.
ServiceController Target internally uses ServiceController Class. But it doesn't return the reason why it couldn't find the service. If you are shure that both computer and service names are correct, the next thing I can suggest to analyze is access violation problems.
And #jsobo's answer can be very useful to diagnose the actual reason because it can show native errors without .Net exception wrappers around them:
sc.exe \Box2 stop MyService