Permalink
Switch branches/tags
Commits on Jun 15, 2018
  1. Further bump version.

    bigpresh committed Jun 15, 2018
    For some reason, there are already git tags for 1.3300 and 1.3301.  I can only
    assume they were trial releases that never actually went anywhere.
    
    Either way, just go for 1.3400.
  2. Update Changes

    bigpresh committed Jun 15, 2018
Commits on Jun 13, 2018
  1. Update Changes

    bigpresh committed Jun 13, 2018
  2. Merge pull request #1189 from PerlDancer/bigpresh/issue_1184_yaml_tes…

    bigpresh committed Jun 13, 2018
    …t_failures
    
    YAML test failures (issue 1184)
  3. Use more-invalid YAML.

    bigpresh committed Jun 13, 2018
    The reason for the failures was that YAML::XS correctly throws an error parsing
    'foo: bar: baz', but YAML.pm doesn't.
    
    Now, there's probably a good argument that YAML.pm *should* baulk at it, but it
    doesn't, and that's not Dancer's fault.
    
    So, use some more obviously-invalid YAML in our test, and all looks good.
  4. Do these YAML-parsing tests against YAML and YAML::XS

    bigpresh committed Jun 13, 2018
    There are some recent test failures, see Issue #1184, which only fail on systems
    without YAML::XS available.
    
    I don't like that these tests used YAML::XS if available and YAML otherwise -
    I'd prefer it to test both, rather than be unpredictable.
    
    So, step 1: test both YAML implementations.
    
    Step 2: find out why it's failing with YAML::XS, and whether it is a big issue.
  5. Merge pull request #1188 from PerlDancer/bigpresh/issue_1187_route_ma…

    bigpresh committed Jun 13, 2018
    …tching
    
    Route matching problems if %+ already populated (issue 1187)
  6. Immediately return if route regex didn't match.

    bigpresh committed Jun 13, 2018
    This is for Issue #1187.
    
    It's not safe to look at `%+` without first having checked whether the regex
    matched successfully or not.  Otherwise, we might see keys from a previous regex
    match, because `%+` is only cleared after a successful match.
    
    (perldoc perlvar confirms, "Note: "%-" and "%+" are tied views into a common
    internal hash associated with the last successful regular expression.")
    
    Dancer2 doesn't have problems here, as it already returns immediately if the
    regex didn't match, the same way this change does.
    
    Note: I think this would only happen via certain deployment methods, and if
    there was an earlier route which used named captures.
    
    In particular, we never had any problems in our large API app's comprehensive
    test suite when we used Dancer::Test.  We've re-worked our testing framework to
    use Plack::Test, partially in readyness to switch to Dancer2, and started
    getting weird failures which were tracked down to incorrect route matching, and
    this was the cause.
  7. bump version

    bigpresh committed Jun 1, 2018
  8. Update Changes

    bigpresh committed Jun 1, 2018
  9. Require MIME::Types 2.17.

    bigpresh committed Jun 1, 2018
    2.16 seems to have some funny ideas:
    
    $ $mt->mimeTypeOf('goatse.zip');
    application/vnd.easykaraoke.cdgdownload
    
    ... what.
Commits on May 23, 2018
  1. Merge pull request #1186 from manwar/add-strict-warnings-pragma

    bigpresh committed May 23, 2018
    Added strict/warnings pragma as reported by CPANTS.
  2. Update Changes

    bigpresh committed May 23, 2018
  3. Bump dist version.

    bigpresh committed May 23, 2018
    (TODO: decide whether to use Dist::Zilla auto-versioning...)
  4. Merge pull request #1185 from PerlDancer/bigpresh/issue_1183_pick_lis…

    bigpresh committed May 23, 2018
    …ten_host_port
    
    issue 1183 - pick listen host port ourselves
Commits on May 22, 2018
  1. Work out which host & port to use ourselves.

    bigpresh committed May 22, 2018
    In PR #1150, I switched to using 127.0.0.10 rather than 127.0.0.1
    for the test server started by Test::TCP to pick a free port on,
    because we were getting some spurious test failures due to a race
    condition - Test::TCP would pick a free port using Net::EmptyPort, but by the
    time the test app had started and tried to bind to it, something else got it
    instead.
    
    That doesn't work on FreeBSD and possibly others, thuough - they only have
    127.0.0.1/32 on the loopback interface by default, not 127/8.
    
    So, work out for ourselves whether to use an IP unlikely to have any ports in
    use if we can, and 127.0.0.1 otherwise, and pick an available port using the
    port 0 trick (request port 0, let the OS pick an available port for us).
    
    TODO: this may need to be made more generic and shared with other tests.
    
    TODO: this test is failing - check if it's down to my work here or something
    else.
    
    TODO right now: go to bed.
  2. More dist.ini & MANIFEST updates.

    bigpresh committed May 22, 2018
    Tweak dist.ini to be closer to previous behaviour.
Commits on May 20, 2018
  1. Update Changes

    bigpresh committed May 20, 2018
Commits on May 18, 2018
  1. Merge pull request #1178 from PerlDancer/bigpresh/unbundle_http_body_…

    bigpresh committed May 18, 2018
    …make_request_body_optional_2
    
    New setting to disable keeping raw request body in RAM
Commits on May 17, 2018
  1. Tests fail listening on 127.0.0.10.

    bigpresh committed May 17, 2018
    This had previously been done in PR #1150 to avoid occasional test failures
    where the port that Test::TCP picked to listen on ended up in use before it
    tried to bind - but that didn't work for *BSD boxes, or, possibly even, any
    non-Linux boxes.
    
    It was reverted, but is still here.  Hmm.
  2. There's one more test now.

    bigpresh committed May 17, 2018
Commits on May 14, 2018
  1. We don't need this, it gets generated by a dzil plugin

    bigpresh committed May 14, 2018
    Dist::Zilla::Plugin::Test::Compile creates this for us, we don't need to have it
    in the repo too, otherwise builds fail with:
    
    ```
    [DZ] attempt to add t/00-compile.t multiple times; added by: filename set by
    GatherDir (Dist::Zilla::Plugin::GatherDir line 225); encoded_content added by
    @Filter/GatherDir (Dist::Zilla::Plugin::GatherDir line 226); content added by
    @Filter/Test::Compile (Dist::Zilla::Plugin::Test::Compile line 169); content set
    by Test::Compile (Dist::Zilla::Plugin::Test::Compile line 205)
    aborting; duplicate files would be produced at
    /home/travis/perl5/perlbrew/perls/5.26.2/lib/site_perl/5.26.2/Dist/Zilla/App/Command/smoke.pm
    line 64.
    ```
  2. Merge branch 'devel' into bigpresh/unbundle_http_body_make_request_bo…

    bigpresh committed May 14, 2018
    …dy_optional_2
    
    Conflicts:
    	dist.ini
  3. Merge pull request #1180 from PerlDancer/bigpresh/use_dzil_dancer_plu…

    bigpresh committed May 14, 2018
    …ginbundle
    
    Use dzil plugin bundle Dancer, not YANICK
Commits on May 13, 2018
  1. Skip Test::NoTabs for now.

    bigpresh committed May 13, 2018
    Several files do contain them:
    https://travis-ci.org/PerlDancer/Dancer/jobs/378209712
    
    ```
    t/author-no-tabs.t .................................. 1/?
    ```
    
    In time I'll fix up all those scripts and then re-add it, but for now, meh.
Commits on May 12, 2018
  1. re-add AutoPrereqs.

    bigpresh committed May 12, 2018
    We don't *have* a cpanfile to use, so let dzil do it for now at least.
  2. Fix merge conflict.

    bigpresh committed May 12, 2018
    (Not quite sure how I ended up committing this with merge markers in...)
  3. dzil tweaks - hardcode version etc.

    bigpresh committed May 12, 2018
    Still not comfortable with Dist::Zilla.
    
    Hardcode the version I want the next release to be.
    
    Extract pre-reqs from cpanfile.
Commits on Mar 26, 2018
Commits on Mar 21, 2018
  1. Test newer perls in travis, and don't require test deps for dzil

    Grinnz authored and bigpresh committed Mar 20, 2018
    As List::Lazy is only a test dep, this should fix the current failure back to 5.14.