Permalink
Commits on Sep 20, 2016
  1. v3.000005

    rjbs committed Sep 20, 2016
            - set up kit_reader and manifest_reader with a helper attribute instead
              of the weird and rarely-seen Moose "initializer" system
  2. gitignore gitnxtver_cache

    rjbs committed Sep 20, 2016
Commits on Sep 19, 2016
  1. use a second attribute, not initializer, for obj-from-class setup

    rjbs committed Sep 19, 2016
    We have some places where we want to be able to have the manifest
    contain a class name, but the attribute contain an object built from
    that.  Because the build object needs to have a reference back to
    the kit, it can't be created entirely from coercion, as types are
    not dependent on the attributes using them.
    
    Instead, we insert a shim attribute that will store the seed of the
    value to use.  There's no danger of someone passing a full object
    into ->new, either, because it couldn't exist yet, because it would
    need to already point back to the instance we're about to construct!
Commits on Aug 4, 2016
  1. v3.000004

    rjbs committed Aug 4, 2016
            - require a newer Email::MIME so we can rely on the header_raw method
  2. remove declared prereqs for AutoPrereqs

    rjbs committed Aug 4, 2016
    ...and require a newer Email::MIME for ->header_raw
Commits on Aug 3, 2016
  1. v3.000003

    rjbs committed Aug 3, 2016
            - manifest.json is now properly decoded while being read, so
              UTF-8 strings in the manifest.json are treated the same way as
              \u.... sequences.  If you had been doubly-UTF-8-encoding your
              manifest.json to work around this bug, your code is now broken
              because my code is now fixed.
  2. correct decoding of UTF-8 in JSON manifests

    rjbs committed Aug 3, 2016
    Inexplicably, I had commented that we definitely did not want to decode
    the octets in JSON.  This statement makes zero sense and I honestly have
    no idea what I thought I was saying.
    
    The UTF-8 read from disk was being passed to JSON.pm as if it was already
    decoded, meaning that UTF-8 on disk would become mojibake when read.  On
    the other hand, \u.... sequences in the JSON would properly become decoded
    text.  I wrote a test to demonstrate this bug, then fixed it.
    
    I assume that this was some kind of cognitive overload from when I was
    fixing the bigger encoding problems in v3, but I do not care to investigate.
Commits on Feb 19, 2015
  1. v3.000002

    rjbs committed Feb 19, 2015
            - immutabilize more classes
  2. Merge branch 'encoded-files'

    rjbs committed Feb 19, 2015
  3. make all classes immutable

    rjbs committed Feb 19, 2015
Commits on Dec 1, 2014
  1. v3.000001

    rjbs committed Dec 1, 2014
              [BREAKING CHANGES]
            - content read from disk (templates, for example) is now decoded
              as UTF-8 if and only if the content type for which it is being used
              is text/*
            - for more information on this change, consult the "ENCODING" section
              of the documentation for Email::MIME::Kit
    
              [BUG FIXES]
            - non-text parts are no longer padded with CRLFs
Commits on Nov 21, 2014
  1. v3.000000

    rjbs committed Nov 21, 2014
              [BREAKING CHANGES]
            - content read from disk (templates, for example) is now decoded
              as UTF-8 if and only if the content type for which it is being used
              is text/*
            - for more information on this change, consult the "ENCODING" section
              of the documentation for Email::MIME::Kit
    
              [BUG FIXES]
            - non-text parts are no longer padded with CRLFs
  2. update changelog

    rjbs committed Nov 21, 2014
Commits on Nov 20, 2014
  1. test for (and fix) breaking of binary attachments

    rjbs committed Nov 20, 2014
    1.  Since we now futz with encoding/decoding more often (and more
        correctly) make sure it doesn't affect binary files.
    
    2.  Realize that even without the encoding, we were breaking binary
        files by adding CRLF even to non-text.
  2. charset UTF-8, not utf-8

    rjbs committed Nov 20, 2014
    They are equivalent, but David Golden was giving me that
    sideeye...
  3. fix test for encoded non-ASCII text content in msg

    rjbs committed Nov 20, 2014
    Gosh, all this time the test has been looking for bogus (Latin-1)
    content instead of UTF-8.
  4. look for non-ASCII, not 8th-bit-set

    rjbs committed Nov 20, 2014
Commits on Nov 2, 2014
  1. v2.102015

    rjbs committed Nov 2, 2014
            - read kit entries in binary mode for the sake of Win32 (thanks,
              Octavian Rasnita!)
Commits on Aug 19, 2014
  1. pep: dead?

    rjbs committed Aug 19, 2014
  2. Remove link to 404ing page

    frioux committed with rjbs Jul 28, 2014
Commits on Jun 6, 2014
  1. v2.102014

    rjbs committed Jun 6, 2014
            - break a circular reference in the assembler (thanks, Charlie
              Garrison)
  2. mark binary files binary

    rjbs committed Jun 6, 2014
  3. make room for VERSION assignments

    rjbs committed Jun 6, 2014
Commits on Dec 31, 2013
Commits on Aug 13, 2013
  1. v2.102013

    rjbs committed Aug 13, 2013
              update bugtracker and repo
    
              only call _set_attachment_info when it is present