Search Results

Search found 1257 results on 51 pages for 'repositories'.

Page 14/51 | < Previous Page | 10 11 12 13 14 15 16 17 18 19 20 21  | Next Page >

  • Jersey 2 in GlassFish 4 - First Java EE 7 Implementation Now Integrated (TOTD #182)

    - by arungupta
    The JAX-RS 2.0 specification released their Early Draft 3 recently. One of my earlier blogs explained as the features were first introduced in the very first draft of the JAX-RS 2.0 specification. Last week was another milestone when the first Java EE 7 specification implementation was added to GlassFish 4 builds. Jakub blogged about Jersey 2 integration in GlassFish 4 builds. Most of the basic functionality is working but EJB, CDI, and Validation are still a TBD. Here is a simple Tip Of The Day (TOTD) sample to get you started with using that functionality. Create a Java EE 6-style Maven project mvn archetype:generate -DarchetypeGroupId=org.codehaus.mojo.archetypes -DarchetypeArtifactId=webapp-javaee6 -DgroupId=example -DartifactId=jersey2-helloworld -DarchetypeVersion=1.5 -DinteractiveMode=false Note, this is still a Java EE 6 archetype, at least for now. Open the project in NetBeans IDE as it makes it much easier to edit/add the files. Add the following <respositories> <repositories> <repository> <id>snapshot-repository.java.net</id> <name>Java.net Snapshot Repository for Maven</name> <url>https://maven.java.net/content/repositories/snapshots/</url> <layout>default</layout> </repository></repositories> Add the following <dependency>s <dependency> <groupId>junit</groupId> <artifactId>junit</artifactId> <version>4.10</version> <scope>test</scope></dependency><dependency> <groupId>javax.ws.rs</groupId> <artifactId>javax.ws.rs-api</artifactId> <version>2.0-m09</version> <scope>test</scope></dependency><dependency> <groupId>org.glassfish.jersey.core</groupId> <artifactId>jersey-client</artifactId> <version>2.0-m05</version> <scope>test</scope></dependency> The complete list of Maven coordinates for Jersey2 are available here. An up-to-date status of Jersey 2 can always be obtained from here. Here is a simple resource class: @Path("movies")public class MoviesResource { @GET @Path("list") public List<Movie> getMovies() { List<Movie> movies = new ArrayList<Movie>(); movies.add(new Movie("Million Dollar Baby", "Hillary Swank")); movies.add(new Movie("Toy Story", "Buzz Light Year")); movies.add(new Movie("Hunger Games", "Jennifer Lawrence")); return movies; }} This resource publishes a list of movies and is accessible at "movies/list" path with HTTP GET. The project is using the standard JAX-RS APIs. Of course, you need the trivial "Movie" and the "Application" class as well. They are available in the downloadable project anyway. Build the project mvn package And deploy to GlassFish 4.0 promoted build 43 (download, unzip, and start as "bin/asadmin start-domain") as asadmin deploy --force=true target/jersey2-helloworld.war Add a simple test case by right-clicking on the MoviesResource class, select "Tools", "Create Tests", and take defaults. Replace the function "testGetMovies" to @Testpublic void testGetMovies() { System.out.println("getMovies"); Client client = ClientFactory.newClient(); List<Movie> movieList = client.target("http://localhost:8080/jersey2-helloworld/webresources/movies/list") .request() .get(new GenericType<List<Movie>>() {}); assertEquals(3, movieList.size());} This test uses the newly defined JAX-RS 2 client APIs to access the RESTful resource. Run the test by giving the command "mvn test" and see the output as ------------------------------------------------------- T E S T S-------------------------------------------------------Running example.MoviesResourceTestgetMoviesTests run: 1, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 0.561 secResults :Tests run: 1, Failures: 0, Errors: 0, Skipped: 0 GlassFish 4 contains Jersey 2 as the JAX-RS implementation. If you want to use Jersey 1.1 functionality, then Martin's blog provide more details on that. All JAX-RS 1.x functionality will be supported using standard APIs anyway. This workaround is only required if Jersey 1.x functionality needs to be accessed. The complete source code explained in this project can be downloaded from here. Here are some pointers to follow JAX-RS 2 Specification Early Draft 3 Latest status on specification (jax-rs-spec.java.net) Latest JAX-RS 2.0 Javadocs Latest status on Jersey (Reference Implementation of JAX-RS 2 - jersey.java.net) Latest Jersey API Javadocs Latest GlassFish 4.0 Promoted Build Follow @gf_jersey Provide feedback on Jersey 2 to [email protected] and JAX-RS specification to [email protected].

    Read the article

  • Migrating Virtual Iron guest to Oracle VM 3.x

    - by scoter
    As stated on the official site, Oracle in 2009, acquired a provider of server virtualization management software named Virtual Iron; you can find all the acquisition details at this link. Into the FAQ on the official site you can also view that, for the future, Oracle plans to fully integrate Virtual Iron technology into Oracle VM products, and any enhancements will be delivered as a part of the combined solution; this is what is going on with Oracle VM 3.x. So, customers started asking us to migrate Virtual Iron guests to Oracle VM. IMPORTANT: This procedure needs a dedicated OVM-Server with no-guests running on top; be careful while execute this procedure on production environments. In these little steps you will find how-to migrate, as fast as possible, your guests between VI ( Virtual Iron ) and Oracle VM; keep in mind that OracleVM has a built-in P2V utility ( Official Documentation )  that you can use to migrate guests between VI and Oracle VM. Concepts: VI repositories.  On VI we have the same "repository" concept as in Oracle VM; the difference between these two products is that VI use a raw-lun as repository ( instead of using ocfs2 and its capabilities, like ref-links ). The VI "raw-lun" repository, with a pure operating-system perspective, may be presented as in this picture: Infact on this "raw-lun" VI create an LVM2 volume-group. The VI "raw-lun" repository, with an hypervisor perspective, may be presented as in this picture: So, the relationships are: LVM2-Volume-Group <-> VI Repository LVM2-Logical-Volume <-> VI guest virtual-disk The first step is to present the VI repository ( raw-lun ) to your dedicated OVM-Server. Prepare dedicated OVM-Server On the OVM-Server ( OVS ) you need to discover new lun and, after that, discover volume-group and logical-volumes containted in VI repository; due to default OVS configuration you need to edit lvm2 configuration file: /etc/lvm/lvm.conf     # By default for OVS we restrict every block device:     # filter = [ "r/.*/" ] and comment the line starting with "filter" as above. Now you have to discover the raw-lun presented and, next, activate volume-group and logical-volumes: #!/bin/bash for HOST in `ls /sys/class/scsi_host`;do echo '- - -' > /sys/class/scsi_host/$HOST/scan; done CPATH=`pwd` cd /dev for DEVICE in `ls sd[a-z] sd?[a-z]`;do echo '1' > /sys/block/$DEVICE/device/rescan; done cd $CPATH cd /dev/mapper for PARTITION in `ls *[a-z] *?[a-z]`;do partprobe /dev/mapper/$PARTITION; done cd $CPATH vgchange -a yAfter that you will see a new device:[root@ovs01 ~]# cd /dev/6000F4B00000000000210135bef64994[root@ovs01 6000F4B00000000000210135bef64994]# ls -l 6000F4B0000000000061013* lrwxrwxrwx 1 root root 77 Oct 29 10:50 6000F4B00000000000610135c3a0b8cb -> /dev/mapper/6000F4B00000000000210135bef64994-6000F4B00000000000610135c3a0b8cb By your OVM-Manager create a guest server with the same definition as on VI:same core number as VI source guestsame memory as VI source guestsame number of disks as VI source guest ( you can create OVS virtual disk with a small size of 1GB because the "clone" will, eventually, extend the size of your new virtual disks )Summarizing:source-virtual-disk path ( VI ):/dev/mapper/6000F4B00000000000210135bef64994-6000F4B00000000000610135c3a0b8cbdest-virtual-disk path ( OVS ):/OVS/Repositories/0004fb00000300006cfeb81c12f12f00/VirtualDisks/0004fb000012000055e0fc4c5c8a35ee.img ** ** = to identify your virtual disk you have verify its name under the "vm.cfg" file of your new guest.Clone VI virtual-disk to OVS virtual-diskdd if=/dev/mapper/6000F4B00000000000210135bef64994-6000F4B00000000000610135c3a0b8cb of=/OVS/Repositories/0004fb00000300006cfeb81c12f12f00/VirtualDisks/0004fb000012000055e0fc4c5c8a35ee.img Clean unsupported parameters and changes on OVS.1. Restore original /etc/lvm/lvm.conf    # By default for OVS we restrict every block device:     filter = [ "r/.*/" ]    and uncomment the line starting with "filter" as above.2. Force-stop lvm2-monitor service  # service lvm2-monitor force-stop 3. Restore original /etc/lvm directories ( archive, backup and cache )  # cd /etc/lvm  # rm -fr archive backup cache; mkdir archive backup cache4. Reboot OVSRefresh OVS repository and start your guest.By OracleVM Manager refresh your repository:By OracleVM Manager start your "migrated" guest: Comments and corrections are welcome.  Simon COTER 

    Read the article

  • Making mercurial subrepositories behave like subversion externals

    - by Emily Dickinson
    Hi guys, The FAQ, and hginit.com have been really useful for helping me make the transition from svn to hg. However, when it comes to using Hg's subrepository feature in the manner of subversion's externals, I've tried everythign and cannot replicate the nice behavior of svn externals. Here's the simplest example of what I want to do: Init "lib" repository This repository is never to be used as a standalone; it's always included by main repositories, as a sub-repository. Init one or more including repositories To keep the example simple, I'll "init" a repository called "main" Have "main" include "lib" as a subrepository Importantly -- AND HERE'S WHAT I CAN'T GET TO WORK: When I modify a file inside of "main/lib", and I push the modification, then that change gets pushed to the "lib" repository -- NOT to a copy inside of "main". Command lines speak louder than words. I've tried so many variations on this theme, but here's the gist. If someone can reply, in command lines, I'll be forever grateful! 1. Init "lib" repository $ cd /home/moi/hgrepos ## Where I'm storing my hg repositories, on my main server $ hg init lib $ echo "foo" lib/lib.txt $ hg add lib $ hg ci -A -m "Init lib" lib 2. Init "main" repository, and include "lib" as a subrepos $ cd /home/moi/hgrepos $ hg init main $ echo "foo" main/main.txt $ hg add main $ cd main $ hg clone ../lib lib $ echo "lib=lib" .hgsub $ hg ci -A -m "Init main" . This all works fine, but when I make a clone of the "main" repository, and make local modifications to files in "main/lib", and push them, the changes get pushed to "main/lib", NOT to "lib". IN COMMAND-LINE-ESE, THIS IS THE PROBLEM: $ /home/moi/hg-test $ hg clone ssh://[email protected]/hgrepos/lib lib $ hg clone ssh://[email protected]/hgrepos/main main $ cd main $ echo foo lib/lib.txt $ hg st M lib.txt $ hg com -m "Modified lib.txt, from inside the main repos" lib.txt $ hg push pushing to ssh://[email protected]/hgrepos/main/lib That last line of output from hg shows the problem. It shows that I've made a modification to a COPY of a file in lib, NOT to a file in the lib repository. If this were working as I'd like it to work, the push would be to hgrepos/lib, NOT to hgrepos/main/lib. I.e., I would see: $ hg push pushing to ssh://[email protected]/hgrepos/lib IF YOU CAN ANSWER THIS IN TERMS OF COMMAND LINES RATHER THAN IN ENGLISH, I WILL BE ETERNALLY GRATEFUL! Thank you in advance! Emily in Portland

    Read the article

  • Best Practices for Source Control Dependencies

    - by VirtuosiMedia
    How do you handle source control setup of a project that has dependency on a separate framework or library? For example, Project A uses Framework B. Should Project A also include the code from Framework B in its repository? Is there a way for it to be included automatically from a different repository or would I have to updated it manually? What are the general approaches are usually taken for this scenario? Assume that I control the repositories for both Project A and Framework B and that the source code for both is not compiled. Any resources or suggestions would be greatly appreciated. I'm currently using Subversion (on a very basic level), but I would like to switch to Mercurial so that I can try out Kiln with Fogbugz. Edit: In Mercurial, would you use parent repositories for this function?

    Read the article

  • AnkhSVN Commits Are Very Slow

    - by jakdep
    Recently, I had to move my SVN repositories to a different server, but I am experiencing some performance problems since the move. I am using Visual Studio 2005, AnkhSVN 2.1.7819.411 and TortoiseSVN 1.6.6 on my workstation and VisualSVN Server on the server which runs Windows Server 2008. Whenever I try to commit a file or view the file history in Visual Studio it takes twenty odd seconds. I confirmed that an exception has been made for VisualSVN Server on the server's firewall, but when I disable the server's firewall the performance is back to normal (1-2 seconds for a commit). When I do a commit or check the log on a file in TortoiseSVN the performance is fine as well. To ensure that the problem was not related to the moving of the repositories, I am running these tests against a new repository which was created on the new server. So, I reckon the problem lies with AnkhSVN, but am at a loss as how to diagnose it further. Any help would be greatly appreciated.

    Read the article

  • C# Design Questions

    - by guazz
    How to approach unit testing of private methods? I have a class that loads Employee data into a database. Here is a sample: public class EmployeeFacade { public Employees EmployeeRepository = new Employees(); public TaxDatas TaxRepository = new TaxDatas(); public Accounts AccountRepository = new Accounts(); //and so on for about 20 more repositories etc. public bool LoadAllEmployeeData(Employee employee) { if (employee == null) throw new Exception("..."); EmployeeRepository emps = new EmployeeRepository(); bool exists = emps.FetchExisting(emps.Id); if (!exists) { emps.AddNew(); } try { emps.Id = employee.Id; emps.Name = employee.EmployeeDetails.PersonalDetails.Active.Names.FirstName; emps.SomeOtherAttribute; } catch() {} try { emps.Save(); } catch(){} try { LoadorUpdateTaxData(employee.TaxData); } catch() {} try { LoadorUpdateAccountData(employee.AccountData); } catch() {} ... etc. for about 20 more other employee objects } private bool LoadorUpdateTaxData(employeeId, TaxData taxData) { if (taxData == null) throw new Exception("..."); ...same format as above but using AccountRepository } private bool LoadorUpdateAccountData(employee.TaxData) { ...same format as above but using TaxRepository } } I am writing an application to take serialised objects(e.g. Employee above) and load the data to the database. I have a few design question that I would like opinions on: A - I am calling this class "EmployeeFacade" because I am (attempting?) to use the facade pattern. Is it good practace to name the pattern on the class name? B - Is it good to call the concrete entities of my DAL layer classes "Repositories" e.g. "EmployeeRepository" ? C - Is using the repositories in this way sensible or should I create a method on the repository itself to take, say, the Employee and then load the data from there e.g. EmployeeRepository.LoadAllEmployeeData(Employee employee)? I am aim for cohesive class and but this will requrie the repository to have knowledge of the Employee object which may not be good? D - Is there any nice way around of not having to check if an object is null at the begining of each method? E - I have a EmployeeRepository, TaxRepository, AccountRepository declared as public for unit testing purpose. These are really private enities but I need to be able to substitute these with stubs so that the won't write to my database(I overload the save() method to do nothing). Is there anyway around this or do I have to expose them? F - How can I test the private methods - or is this done (something tells me it's not)? G- "emps.Name = employee.EmployeeDetails.PersonalDetails.Active.Names.FirstName;" this breaks the Law of Demeter but how do I adjust my objects to abide by the law?

    Read the article

  • Ivy via Nexus proxy

    - by Matthias Hryniszak
    Hi, does anyone knows how do I specify in Ivy something like mirror/mirrorOf in Maven? I'm working with a local Maven proxy (Nexus) and need the tool to specify which of the parent repositories should Nexus proxy be accessing. In Maven I do simply: <mirrors> <mirror> <id>central-mirror</id> <mirrorOf>central</mirrorOf> <url>http://localhost:8081/content/repositories/central</url> </mirror> </mirrors> but I can't find this kind of option in Ivy.

    Read the article

  • AnkhSVN: moving a project to another repo

    - by pcampbell
    My task is to move this VS solution and projects to another SVN server. I'm working with Visual Studio 2010 RC1 and AnkhSVN 2.1.7819.... Currently the files are all bound to a repo at C:\Repositories\foo. I'd like to move it to http://someSite/svn/foo I'm presented with this error message: Repository UUID '152c39db-5799-4234-85f2-074004a6fcad' doesn't match expected UUID '6c83444d-7f93-d64a-b0a0-23283495cf17' Questions How can I avoid this message in AnkhSVN? Are there better solutions for moving the source into a repo on the new target? How can I get Ankh to 'forget' that repo at C:\Repositories\ forever?

    Read the article

  • Linq to SQL DataContext Windsor IoC memory leak problem

    - by Mr. Flibble
    I have an ASP.NET MVC app that creates a Linq2SQL datacontext on a per-web-request basis using Castler Windsor IoC. For some reason that I do not fully understand, every time a new datacontext is created (on every web request) about 8k of memory is taken up and not released - which inevitably causes an OutOfMemory exception. If I force garbage collection the memory is released OK. My datacontext class is very simple: public class DataContextAccessor : IDataContextAccessor { private readonly DataContext dataContext; public DataContextAccessor(string connectionString) { dataContext = new DataContext(connectionString); } public DataContext DataContext { get { return dataContext; } } } The Windsor IoC webconfig to instantiate this looks like so: <component id="DataContextAccessor" service="DomainModel.Repositories.IDataContextAccessor, DomainModel" type="DomainModel.Repositories.DataContextAccessor, DomainModel" lifestyle="PerWebRequest"> <parameters> <connectionString> ... </connectionString> </parameters> </component> Does anyone know what the problem is, and how to fix it?

    Read the article

  • structuremap configuration change based on settings in app.config

    - by user74825
    I am using structuremap in my project. To inject different implementation of a repository, I want to have a switch in app.config which changes all real implementation of a repository to a mock repositories. Lets say IRepository has two implementations RealRepository and MockRepository ForRequestedType() .TheDefaultIsConcreteType(); I want to have a switch in app.config / web.config say (Mock=1), which changes all real repositories implementation to ForRequestedType() .TheDefaultIsConcreteType(); I don't want to write whole plugin definition in app.config, just want one switch, how do I implement this?

    Read the article

  • How to run unit tests with DSSS and GDC?

    - by Benoit Vidis
    I am very new to D and still battling trying to configure my toolchain. I am running Ubuntu Karmic and would like to use DSSS with GDC and Tango or TangoBos. Till now, I installed GDC from Ubuntu repositories, DSSS, Tango and TangoBos from these repositories and I can compile using dsss + gdc + tangobos. According to DSSS documentation, it should be possible to run the unit tests using $ dsss build --test but on my system, the --test argument is ignored. I have dsss last version (0.78) and its inline help does not include anything about unit tests. Running ldc --unittest works fine (though I do not know exactly which libray it picks up). Is there a way to run my unit tests using the same compiler & library than for compilation? If so, is there a way to automate the testing or will I have to run it module per module?

    Read the article

  • Spring online repository for Maven

    - by Fortyrunner
    I've just installed Maven2 for the first time. By default it pulls down a few useful jars into a local project: jakarta-commons, junit etc. I wanted to pull in the latest Spring release (2.5.6 at the time of writing). But the online repositories I looked at (iBiblio and Maven) only had much older versions of Spring libraries. Are there any other repositories that are kept up to date? What is the best practice here; can we maintain them ourselves? I would be prepared to help out maintaining this stuff!

    Read the article

  • git-diff in another directory

    - by ABach
    I'm currently writing a little zsh function that checks all of my git repositories to see if they're dirty or not and then prints out the ones that need a commit. Thus far, I've figured out that the quickest way to figure out a git repository's clean/dirty status is via git-diff and git-ls-files: if ! git diff --quiet || git ls-files --others --exclude-standard; then state=":dirty" fi I have two questions for you folks: Does anyone know of a quicker, more efficient way to check for file changes/additions in a git repo? I want my zsh function to be handed a file path (say ~/Code/git-repos/) and check all of the repositories in it. Is there a way to do without having to cd into each directory and run those commands? Something like git-diff --quiet --git-dir="~/Code/git-repos/..." would be fantastic. Thanks! :)

    Read the article

  • Parsing getopts in bash

    - by ABach
    I've got a bash function that I'm trying to use getopts with and am having some trouble. The function is designed to be called by itself (getch), with an optional -s flag (getch -s), or with an optional string argument afterward (so getch master and getch -s master are both valid). The snippet below is where my problem lies - it isn't the entire function, but it's what I'm focusing on: getch() { if [ "$#" -gt 2 ] || [ "$1" = "-h" ] || [ "$1" = "--help" ]; then echo "Usage: $0 [-s] [branch-name]" >&2 return 1 fi while getopts "s" opt; do echo $opt # This line is here to test how many times we go through the loop case $opt in s) squash=true shift ;; *) ;; esac done } The getch -s master case is where the strangeness happens. The above should spit out s once, but instead, I get this: [user@host:git-repositories/temp]$ getch -s master s s [user@host:git-repositories/temp]$ Why is it parsing the -s opt twice?

    Read the article

  • how to implement unitofwork pattern when using subsonic 2.1(Repository pattern) ?

    - by ROHITH
    I am using subsonic repository pattern(2.1) for asp.net mvc application.In my application,there are many repositories like categoryRepository,Blogrepository etc.Inside each of this repository i am calling subsonic's DB.Select().From()...ExecuteReader() and then loading domain objects from those reader. In the controller action i make multiple calls from these repositories for e.g. List<IBlog> blogs=_blogRepository.GetHottestBlogs(); List<ICategory> categories=_categoryRepository.GetAll(); do i have to implement any unitofwork pattern for this ?.My doubt is that how subsonic performs each operation DB.Update/Insert/Select .Is a TransactionScope is enough for batch update or do i have to use SharedDbConnectionScope to get better performance?

    Read the article

  • Objective-c design advice for use of different data sources, swapping between test and live

    - by user200341
    I'm in the process of designing an application that is part of a larger piece of work, depending on other people to build an API that the app can make use of to retrieve data. While I was thinking about how to setup this project and design the architecture around it, something occurred to me, and I'm sure many people have been in similar situations. Since my work is depending on other people to complete their tasks, and a test server, this slows work down at my end. So the question is: What's the best practice for creating test repositories and classes, implementing them, and not having to depend on altering several places in the code to swap between the test classes and the actual repositories / proper api calls. Contemplate the following scenario: GetDataFromApiCommand *getDataCommand = [[GetDataFromApiCommand alloc]init]; getDataCommand.delegate = self; [getDataCommand getData]; Once the data is available via the API, "GetDataFromApiCommand" could use the actual API, but until then a set of mock data could be returned upon the call of [getDataCommand getData] There might be multiple instances of this, in various places in the code, so replacing all of them wherever they are, is a slow and painful process which inevitably leads to one or two being overlooked. In strongly typed languages we could use dependency injection and just alter one place. In objective-c a factory pattern could be implemented, but is that the best route to go for this? GetDataFromApiCommand *getDataCommand = [GetDataFromApiCommandFactory buildGetDataFromApiCommand]; getDataCommand.delegate = self; [getDataCommand getData]; What is the best practices to achieve this result? Since this would be most useful, even if you have the actual API available, to run tests, or work off-line, the ApiCommands would not necessarily have to be replaced permanently, but the option to select "Do I want to use TestApiCommand or ApiCommand". It is more interesting to have the option to switch between: All commands are test and All command use the live API, rather than selecting them one by one, however that would also be useful to do for testing one or two actual API commands, mixing them with test data. EDIT The way I have chosen to go with this is to use the factory pattern. I set up the factory as follows: @implementation ApiCommandFactory + (ApiCommand *)newApiCommand { // return [[ApiCommand alloc]init]; return [[ApiCommandMock alloc]init]; } @end And anywhere I want to use the ApiCommand class: GetDataFromApiCommand *getDataCommand = [ApiCommandFactory newApiCommand]; When the actual API call is required, the comments can be removed and the mock can be commented out. Using new in the message name implies that who ever uses the factory to get an object, is responsible for releasing it (since we want to avoid autorelease on the iPhone). If additional parameters are required, the factory needs to take these into consideration i.e: [ApiCommandFactory newSecondApiCommand:@"param1"]; This will work quite well with repositories as well.

    Read the article

  • AnkhSVN - How to remove old URLs from list of URLs in the "Open from Subversion" dialog box

    - by user2942597
    I work for a small company and am the sole developer using AnkhSVN to version my code. For the server side I am using VisualSVN v2.5.8. The server is installed on my own machine (on a different drive). I have a few repositories that I created about two years ago that have been working fine. We recently completed an Active Directory Domain Rename (that's another story) so the FQDN of my machine changed so the domain portion is no longer the same as what it was when the server portion was installed. I managed to get AnkhSVN to connect to the repositories so everything is working again but the URL list that comes up on the "Open from Subversion" dialog box still has all the old URLs. How can I remove them? I've searched everywhere I can think of looking for this list but can't seem to find it anywhere. Any suggestions would be greatly appreciated. Chuck R.

    Read the article

  • StructureMap - Injecting a dependency into a base class?

    - by David
    In my domain I have a handful of "processor" classes which hold the bulk of the business logic. Using StructureMap with default conventions, I inject repositories into those classes for their various IO (databases, file system, etc.). For example: public interface IHelloWorldProcessor { string HelloWorld(); } public class HelloWorldProcessor : IHelloWorldProcessor { private IDBRepository _dbRepository; public HelloWorldProcessor(IDBRepository dbRepository) { _dbRepository = dbrepository; } public string HelloWorld(){ return _dbRepository.GetHelloWorld(); } } Now, there are some repositories that I'd like to be available to all processors, so I made a base class like this: public class BaseProcessor { protected ICommonRepository _commonRepository; public BaseProcessor(ICommonRepository commonRepository) { _commonRepository = commonRepository; } } But when my other processors inherit from it, I get a compiler error on each one saying that there's no constructor for BaseProcessor which takes zero arguments. Is there a way to do what I'm trying to do here? That is, to have common dependencies injected into a base class that my other classes can use without having to write the injections into each one?

    Read the article

  • Composer does not find dependencies of vcs repository

    - by Michael Freund
    i've got a strange problem ... project-a is my main project. project-b is my library, checked in to subversion composer.json of project-b { "name": "fragger/baseclasses", "version" : "0.0.1-dev", "description": "Baseclasses and Interfaces", "require": { "silex/silex": "1.0.x-dev", "3rd-party/smarty": "3.*", "swiftmailer/swiftmailer": "4.2-dev" }, "autoload": { "psr-0": { "baseclasses": "src/" } } } and composer.json of project-b { "repositories" : [ { "type": "vcs", "url" : "svn+ssh://....." } ], "require": { "fragger/baseclasses": ">=0.0.1-dev" } } output of install command php composer.phar install Loading composer repositories with package information Installing dependencies Your requirements could not be resolved to an installable set of packages. Problem 1 - Installation request for fragger/baseclasses >=0.0.1-dev -> satisfiable by fragger/baseclasses dev-trunk. - fragger/baseclasses dev-trunk requires silex/silex 1.0.x-dev -> no matching package found. But a composer install in project a alone, works fine

    Read the article

  • Mercurial local repository backup

    - by Ricket
    I'm a big fan of backing things up. I keep my important school essays and such in a folder of my Dropbox. I make sure that all of my photos are duplicated to an external drive. I have a home server where I keep important files mirrored across two drives inside the server (like a software RAID 1). So for my code, I have always used Subversion to back it up. I keep the trunk folder with a stable copy of my application, but then I create a branch named with my username, and inside there is my working copy. I make very few changes between commits to that branch, with the understanding that the code in there is my backup. Now I'm looking into Mercurial, and I must admit I haven't truly used it yet so I may have this all wrong. But it seems to me that you have a server-side repository, and then you clone it to a working directory in the form of a local repository. Then as you work on something, you make commits to that local repository, and when things are in a state to be shared with others, you hg push to the parent repository on the server. Between pushes of stable, tested, bug-free code, where is the backup? After doing some thinking, I've come to the conclusion that it is not meant for backup purposes and it assumes you've handled that on your own. I guess I need to keep my Mercurial local repositories in my dropbox or some other backed-up location, since my in-progress code is not pushed to the server. Is this pretty much it, or have I missed something? If you use Mercurial, how do you backup your local repositories? If you had turned on your computer this morning and your hard drive went up in flames (or, more likely, the read head went bad, or the OS corrupted itself, ...), what would be lost? If you spent the past week developing a module, writing test cases for it, documenting and commenting it, and then a virus wipes your local repository away, isn't that the only copy? So then on the flip side, do you create a remote repository for every local repository and push to it all the time? How do you find a balance? How do you ensure your code is backed up? Where is the line between using Mercurial as backup, and using a local filesystem backup utility to keep your local repositories safe?

    Read the article

  • VB.NET template instance - passing a variable data type

    - by FerretallicA
    As the title suggests, I'm tyring to pass a variable data type to a template class. Something like this: frmExample = New LookupForm(Of Models.MyClass) 'Works fine Dim SelectedType As Type = InstanceOfMyClass.GetType() 'Works fine repoGeneric = New Repositories.Repository(Of SelectedType) 'Ba-bow! repoGeneric = New Repositories.Repository(Of InstanceOfMyClass.GetType()) 'Ba-bow! I'm assuming it's something to do with the template being processed at compile time but even if I'm off the mark there, it wouldn't solve my problem anyway. I can't find any relevant information on using Reflection to instance template classes either. (How) can I create an instance of a dynamically typed repository at runtime?

    Read the article

  • How do I solve "405 Method Not Allowed" for our subversion setup?

    - by macke
    We're serving our source code using VisualSVN running on Windows Server 2003. Recently, we split a portion of a project into a new project in it's own repository, and then linked it back to the original project using svn:externals. Since then, we've been having issues when we try to commit files with Subclipse. The error we're getting is: svn: Commit failed (details follow): svn: PROPFIND of '/svn': 405 Method Not Allowed (https://svn.ourserver.com) Googling for a while didn't really help, our config seems to be correct. It should also be noted that we've been running this server for a while no without these problems and apart from splitting the project into two repositories, no changes have been made to the server (ie, config files are the same). It should also be noted that these errors only appear when we try to check in multiple files at once. If we check in one file at a time there are no errors. Also, it only appears in Subclipse as far as we know right now, Versions.app (OS X) seems to work fine so that is our current workaround. So, the questions is how do I analyze the error to find the cause and subsequently fix it? I'm by no means a svn guru and right now I'm clueless. EDIT: It seems we can check in multiple files in the same package, but not files from multiple packages. Also, when I "split" the project into two repositories, I imported the original repository with a new name. I did not do a dump and then import that dump. Could that be the source of our issues, and if so, how would I solve that? EDIT: After some jerking around it seems as though it is indeed related to when checking in files in different repositories. If I try to do a single commit in both Repo A and Repo B (referenced by svn:externals) at the same time, I get the error. Versions.app handles this correctly, but I guess it might just be doing two commits, not a single one. Subclipse fails miserably. For now, we simply do multiple commits, one for Repo A and one for Repo B, that works just fine. If anyone smarter than me could fill in the details why this is happening, whether or not this kind of setup is stupid etc, please go right ahead.

    Read the article

  • Best way to version control a WCF application with Git?

    - by Sam
    Suppose I have the following projects. The format is [ProjectName] : [ProjectDependency1, ProjectDependency2, etc.] // Service CoolLibrary WcfApp.Core WcfApp.Contracts WcfApp.Services : CoolLibrary, WcfApp.Core, WcfApp.Contracts // Clients CustomerX.App : WcfApp.Contracts CustomerY.App : WcfApp.Contracts CustomerZ.App : WcfApp.Contracts (On a side note, WcfApp.Contracts should not depend on WcfApp.Core, right? Else CustomerX.App would also depend on and thus be exposed to the service domain model?) (CoolLibrary is shared with other applications, so I can't just put it inside of WcfApp.Services.) All of this code is in-house. I was thinking of having 6 repositories for this. The format is [repository folder name] : [Projects included in repository.] 1. CoolLibrary.git : CoolLibrary 2. WcfApp.Contracts.git : WcfApp.Contracts 3. WcfApp.git : WcfApp.Core, WcfApp.Services 4. CustomerX.App.git : CustomerX.App 5. CustomerY.App.git : CustomerY.App 6. CustomerZ.App.git : CustomerZ.App How should I manage my project dependencies? I see three options: I could use binaries which I have to manually copy to each dependent repository. This would be easiest at the start, but my repositories would be a little bloated, and it'd become more tedious as I add more client apps for customers. I could import dependent code as submodules. This is what I will probably end up doing, although I keep reading on the web that submodules are a hassle. I also read that I can use something called the subtree merge strategy, but I am not sure how it is different from just cloning the repo into a subdirectory and adding the subdirectory to .gitignore. Is the difference that the subtree is recorded in the master repository, so (for example) cloning it from a different location will also pull the subtree? I know I asked a lot of questions in this post, but the most important two questions I have are: 1. Am I using the right number and layout of repositories? Should I use less or more? 2. Which of the three dependency management strategies would you recommend? Is there another strategy I haven't considered?

    Read the article

  • What git gotchas have you been caught by?

    - by Bob Aman
    The worst one I've been caught by was with git submodules. I had a submodule for a project on github. The project was unmaintained, and I wanted to submit patches, but couldn't, so I forked. Now the submodule was pointing at the original library, and I needed it to point at the fork instead. So I deleted the old submodule and replaced it with a submodule for the new project in the same commit. Turns out that this broke everyone else's repositories. I'm still not sure what the correct way of handling this situation is, but I ended up deleting the submodule, having everyone pull and update, and then I created the new submodule, and had everyone pull and update again. It took the better portion of a day to figure that out. What have other people done to accidentally screw up git repositories in non-obvious ways, and how did you resolve it?

    Read the article

< Previous Page | 10 11 12 13 14 15 16 17 18 19 20 21  | Next Page >