Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
tag: nexus-2.4.0-09
Commits on Apr 10, 2013
  1. nexus oss 2.4.0-09

    Sonatype Release Machine authored
    Signed-off-by: Sonatype Release Machine <release@sonatype.com>
Commits on Apr 4, 2013
  1. nexus oss 2.4.0-08

    Sonatype Release Machine authored
    Signed-off-by: Sonatype Release Machine <release@sonatype.com>
Commits on Apr 3, 2013
  1. @ifedorenko

    NXCM-5192 fixed NPE when proxy remote has autorouting disabled

    ifedorenko authored
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
  2. @jdillon

    Merge pull request #786 from sonatype/httpclient-logging-tidy

    jdillon authored
    TRIVIAL: clean up httpclient logging
Commits on Apr 2, 2013
  1. @ifedorenko

    simple script to set version for all projects

    ifedorenko authored
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
  2. @ifedorenko

    version 2.5.0-SNAPSHOT

    ifedorenko authored
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
  3. @ifedorenko

    Cleanup: more WhiteList->Autorouting renames

    ifedorenko authored
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
  4. @ifedorenko

    Fixed IllegalStateException during nexus shutdown

    ifedorenko authored
    ... sorry, Tamas, I was wrong and you were right about evenbus api.
    
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
Commits on Apr 1, 2013
  1. back to version 2.4-SNAPSHOT

    Sonatype Release Machine authored
  2. nexus oss 2.4.0-07

    Sonatype Release Machine authored
    Signed-off-by: Sonatype Release Machine <release@sonatype.com>
  3. forge parent 34

    Sonatype Release Machine authored
  4. @ifedorenko

    Cleanup: renamed and documented purpose of ConstrainedExecutor

    ifedorenko authored
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
  5. @ifedorenko

    NEXUS-5654 fixed race between offer/revoke and initializePrefixFile

    ifedorenko authored
    The race was only possible if a new repository was provisioned and
    deploys to the new repository happened very close in time. Extremely
    unlikely unless repository provisioning and deployment are executed
    by the same script.
    
    The fix is to remove redundant 'unpublish' call during
    initializePrefixFile.
    
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
  6. @ifedorenko

    NEXUS-5654 fixed endless autorouting initialization restart

    ifedorenko authored
    Hosted repositories autorouting initialization involves deep walk
    of entire repository local storage and is likely to take significant
    time for large nexus installations like OSSRH. In attempt to prevent
    lost-update race between artifact deploy/delete and hosted repositry
    autorouting initiazation, deploy/delete unconditionally restarted
    initization which could lead to apparently endless initialization if
    time between deployment is less than initiazation time (again, not
    impossible for larger nexus installations).
    
    The fix is to never restart initialization logic during artifact
    deploy. This is not bulletproof and can result in missing prefix
    entries if deployment of new top-level groupId happens after certain
    point during repository autorouting initiazation, but I believe
    this is going to be very unlikely (assumption is that new top
    groupIds are very rare for populated hosted repositories and
    initiazation is very quick for new/empty hosted repositories). If
    this does happen, the workaround is to delete prefix file and
    wait for 1hr for another initilization attempt to take place.
    
    In order to completely eliminate possibility of race between
    artifact deploy and autorouting initiazation we'd need to extend
    RepositoryItemUidLock API to allow lock-with-timeout, something
    I decided was too tricky and risky this late in 2.4 cycle.
    
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
  7. @ifedorenko

    Cleanup: removed unused method parameter

    ifedorenko authored
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
  8. @ifedorenko

    Cleanup: reconciled autorouting startup and periodic update logic

    ifedorenko authored
    Automatic routing periodic update logic is now able to initialize
    hosted repository prefix path list as needed. No need to do this
    explicitly during startup any more.
    
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
  9. @ifedorenko

    Fixed wrong RoutingStatusResourceTest initialization sequence

    ifedorenko authored
    The test was waiting for routing background updates to finish
    way too early (before nexus repo manager was started) and was
    extremely sensitive to changes auto-routing implementation.
    
    Changed the test to do waiting later during nexus instance
    startup. This is still a workaround and the true fix should
    wait for a particular event, "repo auto-routing publish state
    change" in this case, not for "no backgroup update jobs are
    running" as it does now.
    
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
  10. @ifedorenko

    Cleanup: removed separate no-scrape marker file

    ifedorenko authored
    Changed prefix file format to allow encoding of no-scrape request
    using "@ unsupported" directive. Single automatic routing medata
    file makes proper synchronization easier to implement and simplifies
    overall implementation.
    
    This also eliminated two HEAD requests per remote (due to a bug
    original implementation performed HEAD no-scrape file for each
    discovery strategy, not once per remote).
    
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
Commits on Mar 30, 2013
  1. @ifedorenko

    Cleanup: removed extra prefix and noscrape paths

    ifedorenko authored
    There is currently no use case for these extra file paths,
    they complicate the code and make proper synchronization
    harder to implement.
    
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
  2. @ifedorenko

    NEXUS-5654 fixed deadlock during concurrent deploy

    ifedorenko authored
    Propagation of prefix file changes from members to group repositories
    was acquiring read-lock on all group members while holding write-lock
    on one of the members. Two such changes executed concurrently for
    different members of the same group will almost certainly result in
    deadlock. Solution is to execution group prefix file update on a
    separate thread.
    
    Signed-off-by: Igor Fedorenko <igor@ifedorenko.com>
Commits on Mar 29, 2013
  1. @cstamas

    Fixing the build.

    cstamas authored
    Prefix updates happens async on config
    save, and the UT did not properly wait for
    them.
  2. @cstamas

    Merge pull request #792 from sonatype/nexus-5602-using-updater-for-in…

    cstamas authored
    …itial-update
    
    REVIEW: NEXUS-5602: Using the periodic update thread for initial updates too
Commits on Mar 28, 2013
  1. @cstamas

    NEXUS-5602: Using the periodic update thread for initial updates too

    cstamas authored
    On 1st boot of updated instances, a lot of MavenRepository
    instances (hosted and proxy) will be piled up in the array
    needUpdateRepositories in WLManagerImpl#startup().
    
    Currently, as the thread pool with fixed size of 5 and
    blocking queue was deferring update.
    
    With this change, the for-loop doing initial WL build
    (walk of the content for hosted and discovery for proxies)
    is deferred, and periodic updater thread is used.
    
    This also means, that upon boot, the initial delay of
    first run will be applied, and on that moment on, hosted
    reposes will have WL, not before that.
    
    Initial change breaks all the UTs and ITs that uses
    exposed-for-tests-only method to reach a state when
    the test should start.
    
    isUpdateWhitelistJobRunning() method used for this
    is now updated, and will flag (and make tests wait)
    as long as boot sequence is done (newly added) AND
    there are no bg updates running (as before).
    
    Also, EventDispatcher simplified, the feature active
    flag is not passed anymore, instead, it is not
    registered to event bus when feature is not
    active, resulting in same behavior (before, when
    feature was inactive EventDispatcher got events
    but simply did not invoke WL Manager).
  2. @ifedorenko
  3. @cstamas

    Merge pull request #795 from sonatype/nexus-5639-rename-wl-deep

    cstamas authored
    [blocker 2.4] NEXUS-5639: Deep rename of WL into Routing
  4. @cstamas

    Merge pull request #794 from sonatype/nexus-5639-rename-wl

    cstamas authored
    [2.4 blocker] NEXUS-5639: Rename WL
  5. @cstamas

    NEXUS-5639: Rename thread names

    cstamas authored
    From "wl" to "ar" as Auto routing...
  6. @cstamas

    NEXUS-5639: M2 repo external config change

    cstamas authored
    Renamed externalConfiguration nodes too
  7. @cstamas

    NEXUS-5639: URL change of the resources

    cstamas authored
    Client is handled too
  8. @cstamas

    NEXUS-5639: "Deep" rename WL to Routing

    cstamas authored
    Initial code changes, bundle smoke tested and is working.
    
    Tests and ITs not verified yet.
  9. @cstamas

    NEXUS-5639: Cleaning log messages

    cstamas authored
    In AbstractMavenRepo
  10. @cstamas

    NEXUS-5639: Cleaning log messages

    cstamas authored
    In WLManagerImpl
  11. @cstamas

    NEXUS-5639: Cleaning log messages

    cstamas authored
    In EventDispatcher
  12. @cstamas

    NEXUS-5639: Rename WL

    cstamas authored
    UI panel mods:
    * tab title renamed from "Whitelist" to "Routing"
    * tab name changed from "whitelist" to "routing"
    * "Force remote discovery" button renamed to "Update now"
Commits on Mar 26, 2013
  1. nexus oss 2.4.0-06

    Sonatype Release Machine authored
    Signed-off-by: Sonatype Release Machine <release@sonatype.com>
Something went wrong with that request. Please try again.