Skip to content


Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Commits on May 26, 2015
  1. @tenderlove

    -I should always win over gems

    tenderlove authored
    If there were gems in the "unresolved" list, then the files in that gem
    would be preferred over files that included in -I.  For example, if you
    had a depdedncy tree that looks like this:
    digraph g {
      a -> b [label="version: = 1"];
      b -> c [label="version: > 0"];
    and the installed gems were like this:
    a = 1,
    b = 1,
    c = 1,
    c = 2,
    When the user requires 'a', then 'b' will be immediately activated
    because there is only one decision to make.  However, 'c' will be added
    to the `unresloved_deps` list.  Having anything in that list will cause
    Rubygems [to skip checking if Ruby's `require` can handle the file](
    If the user provides any paths to `-I`
    that contain files that are also in an activated gem (like `b` in this
    case), then Rubygems will prefer the file from the gem rather than the
    file provided by -I.
    This commit fixes it so that -I will win over gems.  e2935c8 introduced
    the incorrect behavior.
Commits on May 24, 2015
  1. @tenderlove

    Merge pull request #1241 from rubygems/rakefile_instructions

    tenderlove authored
    At least explain what to do instead of a LoadError
Commits on May 22, 2015
  1. @djberg96

    Added entry for PR #1250.

    djberg96 authored
  2. @djberg96

    Merge pull request #1250 from pjump/stringio

    djberg96 authored
    remove util/stringio.rb, readd stringio
  3. @djberg96

    Added entry for PR #1253.

    djberg96 authored
  4. @djberg96

    Merge pull request #1253 from amerov/patch-1

    djberg96 authored
    Use Array#compact before Array#uniq
Commits on May 18, 2015
  1. @evanphx
Commits on May 17, 2015
  1. @amerov
Commits on May 15, 2015
  1. @indirect

    Merge pull request #1251 from segiddins/seg-fix-indexer-spelling

    indirect authored
    [Indexer] Finish s/inidicies/indices/ which was breaking the build
  2. @segiddins
  3. @hsbt
  4. @hsbt
  5. @hsbt
  6. @evanphx

    Fix version number affected

    evanphx authored
  7. @evanphx

    Fix new version number

    evanphx authored
Commits on May 14, 2015
  1. @evanphx
  2. @evanphx
  3. @evanphx

    Limit API endpoint to original security domain

    evanphx authored
Commits on May 13, 2015
  1. @pjump

    remove util/stringio.rb, readd stringio

    pjump authored
    the in-ruby emulation of stringio makes Gunzip almost 300 times slower
    (on ruby 2.2.2)
    as compared to using standard libary's stringio
    Gem.gunzip('specs.4.8.gz')) takes about 16 seconds with
    and about 50ms without it
Commits on May 11, 2015
  1. @tenderlove

    implement `Gem.find_files` using stubs instead of full specs

    tenderlove authored
    this speeds up `Gem.find_files` since it doesn't need to instantiate
    specs in order to find files
  2. @tenderlove

    move `matches_for_glob` up to `BasicSpecification`

    tenderlove authored
    `matches_for_glob` doesn't depend on anything in the real specification,
    so we can push it up to the `BasicSpecification`.
  3. @tenderlove

    move `lib_dirs_glob` in to BasicSpecification

    tenderlove authored
    The `lib_dirs_glob` method doesn't depend on anything in the full
    specification, so we can safely push it up to the basic specification
Commits on May 4, 2015
  1. @tenderlove

    change `try_activate` to loop through gemspecs once

    tenderlove authored
    before this commit, `try_activate` would look through all specs twice in
    the worst case.  IRB happens to always hit the worst case when it looks
    up locales.  This change just loops through all specs once.
  2. @tenderlove

    `find_by_path` should search through stubs first

    tenderlove authored
    if `try_activate` fails to find an inactive spec, it will look through
    the active specs via `find_by_path`.  `find_by_path` could use the stubs
    which would be faster in that case.
  3. @tenderlove

    add a test for d3b0914

    tenderlove authored
    reverting d3b0914 would not cause any tests to fail, so this commit
    adds a test surrounding that behavior
Commits on May 3, 2015
  1. @andremedeiros
  2. @indirect
Commits on May 1, 2015
  1. @tenderlove

    fix test warnings

    tenderlove authored
    I accidentally introduced a shadowed variable warning, so fix that.
    Also add a missing require so the test will run in isolation.
  2. @tenderlove

    sort grouped stubs by version

    tenderlove authored
    specs grouped by name have the same name, so avoid calling _resort! and
    just sort by version
  3. @indirect

    Merge pull request #1239 from segiddins/seg-all-resets-specs-by-name

    indirect authored
    [Specification] Also reset specs_by_name when setting all
  4. @segiddins
  5. @segiddins
  6. @segiddins
Commits on Apr 30, 2015
  1. @tenderlove

    change Gem::StubSpecification to remove cache

    tenderlove authored
    StubSpecification should not cache data it doesn't own.  Instead lets
    move the cache to the Stubline class so that it can control it's own
    cache.  The attributes that are cached in Stubline are also cached in
    Gem::Specification.  If `data` happened to be a Gem::Specification, then
    the data would actually be cached twice (once in Gem::Specification,
    then again in StubSpecification).  Lets keep the cache limited
  2. @tenderlove
Something went wrong with that request. Please try again.