Commits on Dec 15, 2010
  1. add unit test for 7defd23

    committed Dec 15, 2010
Commits on Dec 13, 2010
  1. Remove empty string from splited pairs to avoid ugly warnings.

    This fix is for some weird cookie string like '__utmv=XXXXXXX.|4=k4=false=1,;' that is used Google Analytics.
    committed Dec 13, 2010
Commits on Dec 10, 2010
  1. @miyagawa

    Checking in changes prior to tagging of version 0.9956.

    Changelog diff is:
    
    diff --git a/Changes b/Changes
    index 78fefe0..4736441 100644
    --- a/Changes
    +++ b/Changes
    @@ -2,6 +2,15 @@ Revision history for Perl extension Plack
    
     Take a look at http://github.com/miyagawa/Plack/issues for the planned changes before 1.0 release.
    
    +0.9956  Thu Dec  9 19:32:46 PST 2010
    +        - FastCGI: Fixed an empty PATH_INFO with mod_fastcgi (and possibly others)
    +        - FastCGI: Improved the automatic detection of the case when invoked from web server. #141
    +          (reported by LeoNerd)
    +        - plackup: Document that -e 'enable ...' doesn't assume app.psgi when there's no argument. #106 (clkao)
    +        - Plack::App::FCGIDispatcher: Remove the Status: header #123 (reported by Htbaa)
    +        - Apache2: Work around issues where SCRIPT_NAME gets wrong when LocationMatch is used. #136
    +          (reported by atiking)
    +
     0.9955  Thu Dec  9 18:02:50 PST 2010
             - More fixes to a possible directory traversal
    miyagawa committed Dec 9, 2010
  2. @miyagawa

    Figure out when Apache2 handler is running with LocationMatch block. …

    …Fixes gh-136
    
    This is stolen from Catalyst::Engine::Apache where it checks if
    $r->location contains non-URI regular character and in that case treat
    it as a regular expression to figure out the SCRIPT_NAME part. Totally
    janky but there seems no other way to achieve this.
    miyagawa committed Dec 9, 2010
  3. @miyagawa

    Remove the requirement of Apache::Const stuff in apache tests

    This gets in the way when I need to run these .t with non-system perl
    but with httpd that has mod_perl linked against different versions of
    perl. These tests won't run without TEST_APACHE* env var anyway.
    miyagawa committed Dec 9, 2010
  4. @miyagawa
  5. @miyagawa

    Removed the Status header from external FastCGI server. Fixes gh-123

    This is to avoid duplicated Status: header when running FCGIDispatcher
    app itself using FastCGI with mod_fastcgi etc.
    miyagawa committed Dec 9, 2010
  6. @miyagawa

    Detect STDIN being socket earlier in FastCGI handler Fixes gh-141

    Plack::Runner sets port=5000 by default which makes FastCGI handler
    not DWIM when kicked from a web serevr like mod_fastcgi. Now fixed to
    check if -S STDIN earlier.
    miyagawa committed Dec 9, 2010
  7. @miyagawa

    mod_fastcgi doesn't appear to set PATH_INFO when you access the root …

    …without the trailing slash
    miyagawa committed Dec 9, 2010
  8. @miyagawa

    Checking in changes prior to tagging of version 0.9955.

    Changelog diff is:
    
    diff --git a/Changes b/Changes
    index c756bd2..78fefe0 100644
    --- a/Changes
    +++ b/Changes
    @@ -2,6 +2,9 @@ Revision history for Perl extension Plack
    
     Take a look at http://github.com/miyagawa/Plack/issues for the planned changes before 1.0 release.
    
    +0.9955  Thu Dec  9 18:02:50 PST 2010
    +        - More fixes to a possible directory traversal
    +
     0.9954  Thu Dec  9 17:45:59 PST 2010
             - Fixed a directory traversal bug in Plack::App::File etc. RT:63020
    miyagawa committed Dec 9, 2010
  9. @miyagawa
  10. @miyagawa

    Checking in changes prior to tagging of version 0.9954.

    Changelog diff is:
    
    diff --git a/Changes b/Changes
    index 45dbdf0..c756bd2 100644
    --- a/Changes
    +++ b/Changes
    @@ -2,6 +2,9 @@ Revision history for Perl extension Plack
    
     Take a look at http://github.com/miyagawa/Plack/issues for the planned changes before 1.0 release.
    
    +0.9954  Thu Dec  9 17:45:59 PST 2010
    +        - Fixed a directory traversal bug in Plack::App::File etc. RT:63020
    +
     0.9953  Fri Dec  3 14:50:09 PST 2010
             - Include the original error message in the StackTrace text output on console.
               This requires Devel::StackTrace 1.23 and Devel::StackTrace::WithLexicals 0.08 (optional)
    miyagawa committed Dec 9, 2010
  11. @miyagawa

    Fixes the directory traversal bug RT #63020

    Also fixes the false positive bug that catches legit path that ends in ../
    miyagawa committed Dec 9, 2010
Commits on Dec 4, 2010
  1. @miyagawa

    Fixed errors in Changes

    miyagawa committed Dec 3, 2010
Commits on Dec 3, 2010
  1. @miyagawa

    Checking in changes prior to tagging of version 0.9953.

    Changelog diff is:
    
    diff --git a/Changes b/Changes
    index c9773d3..c67bf27 100644
    --- a/Changes
    +++ b/Changes
    @@ -2,6 +2,11 @@ Revision history for Perl extension Plack
    
     Take a look at http://github.com/miyagawa/Plack/issues for the planned changes before 1.0 release.
    
    +0.9953  Fri Dec  3 14:50:09 PST 2010
    +        - Include the original error message in the StackTrace text output on console.
    +          This requires Devel::StackTrace 1.23 and Devel::StackTrace::WithLexicals 0.08 (optional)
    +        - Fixed AccessLog middleware to handle multiple dashes in %{} (Jiro)
    +
     0.9952  Thu Dec  2 14:03:48 PST 2010
             - Fixed the potential deadlocks in WrapCGI's read/write pipe (typester)
             - Improved documentations on plackup -e
    miyagawa committed Dec 3, 2010
  2. @spiritloose @miyagawa
  3. @miyagawa
  4. @miyagawa

    Include original error message in the text version of StackTrace

    Also passes indent => 1 so we don't need the janky trace_as_string().
    Note that this code does not seem to work with Devel::StackTrace::WithLexicals
    as of this writing because it ignores 'message' and 'indent' :(
    
    Going to patch it.
    miyagawa committed Dec 3, 2010
Commits on Dec 2, 2010
  1. @miyagawa

    Checking in changes prior to tagging of version 0.9952.

    Changelog diff is:
    
    diff --git a/Changes b/Changes
    index 77ab5e7..c9773d3 100644
    --- a/Changes
    +++ b/Changes
    @@ -2,6 +2,12 @@ Revision history for Perl extension Plack
    
     Take a look at http://github.com/miyagawa/Plack/issues for the planned changes before 1.0 release.
    
    +0.9952  Thu Dec  2 14:03:48 PST 2010
    +        - Fixed the potential deadlocks in WrapCGI's read/write pipe (typester)
    +        - Improved documentations on plackup -e
    +        - Fixed a potential DoS vulnerability in HTTP::Server::PSGI (kazuho)
    +        - Allows setting names of FCGI process with proc_title option (rafl)
    +
     0.9951  Mon Oct 25 13:50:33 PDT 2010
             - Added Feersum to the benchmark script (stash)
             - Lint: fixed the body handle check to see if the file has getline() method (tokuhirom)
    miyagawa committed Dec 2, 2010
  2. @miyagawa

    Revert StackTrace improvements.

    This series of patches just seems like a workaround and doesn't look
    ideal to me. We could discuss more to merge more appropriate fix to
    the fundamental problem.
    miyagawa committed Dec 2, 2010
  3. @rafl @miyagawa
Commits on Nov 29, 2010
  1. @miyagawa
Commits on Nov 24, 2010
  1. @miyagawa
Commits on Nov 14, 2010
Commits on Nov 7, 2010
  1. @miyagawa
Commits on Nov 2, 2010
  1. @miyagawa

    s/require/use/

    miyagawa committed Nov 2, 2010
  2. @miyagawa
  3. @theory
  4. @theory

    Look for `stack_trace`, too.

    This is to support StackTrace::Auto, which is used by Throwable::X and pernaps other exception classes.
    theory committed Nov 2, 2010
Commits on Nov 1, 2010
  1. @theory

    Improve text representation of stack trace.

    There were a few issues when the original exception was an Exception::Class
    object:
    
    * Ghe text representation was not showing the original error message.
      Devel::StackTrace::AsHTML nicely adds the original exception text to the
      beginning of the HTML, but the text version did not. Instead the trace just
      showed the Exception::Class object getting passed around.
    
    * The stack trace was not the original stack trace, but a new one.
    
    So change the exception handler to see if the exception object is blessed and
    has a `trace` method. If so, use that for the trace. This assumes any
    exception object with a `trace` method returns a Devel::StackTrace object,
    which I think is likely.
    
    Then, change the way the message is displayed to add the original error
    message, as AsHTML does, and to use Devel::StackTrace's `as_string()` method
    rather than our own function that does the same thing.
    theory committed Nov 1, 2010
Commits on Oct 25, 2010
  1. @miyagawa

    Checking in changes prior to tagging of version 0.9951.

    Changelog diff is:
    
    diff --git a/Changes b/Changes
    index e595c95..77ab5e7 100644
    --- a/Changes
    +++ b/Changes
    @@ -2,6 +2,14 @@ Revision history for Perl extension Plack
    
     Take a look at http://github.com/miyagawa/Plack/issues for the planned changes before 1.0 release.
    
    +0.9951  Mon Oct 25 13:50:33 PDT 2010
    +        - Added Feersum to the benchmark script (stash)
    +        - Lint: fixed the body handle check to see if the file has getline() method (tokuhirom)
    +        - StackTrace: store the stacktrace in $env->{'plack.stacktrace.text'} and $env->{'plack.stacktrace.html'} (theory)
    +        - Added ->mount method to the Plack::Builder OO interface (franckcuny)
    +        - HTTPExceptions: Don't set an invalid Content-Length when the exception is not an object (ask)
    +        - ErrorDocument: Fixed wrong Content-Length header be set (ask)
    +
     0.9950  Thu Sep 30 14:11:33 PDT 2010
             - Fixed typos in middleware docs (miyagawa, theory, tokuhirom)
             - App::Directory: fixed URL generation escape bug (chiba)
    miyagawa committed Oct 25, 2010
  2. @miyagawa

    Fixed UUV warnings

    miyagawa committed Oct 25, 2010
  3. @miyagawa

    Storing StackTrace object in $env causes a circular reference.

    The "real" fix for this would be to have an option or method to walk
    over StackTrace frame and weaken() all args and lexicals (in case of
    WithLexicals) so that the trace object can be safely stored in $env,
    but for now, storing text version of the trace and HTML version would
    be enough.
    miyagawa committed Oct 25, 2010
  4. @abh @miyagawa