Skip to content
Permalink
Branch: master
Commits on Mar 15, 2019
  1. Update comment in stdio_changegroup_command

    glandium committed Mar 15, 2019
    188c53d added support for bundle2. It's about time to update this.
  2. Revert "[store-changegroup] Read the whole bundle first"

    glandium committed Mar 15, 2019
    This reverts commit 37cf073.
    
    It turns out there's way too much data at this level when dealing with e.g.
    mozilla-unified. For instance, for mozilla-unified, there's > 7.3GB of
    chunks to deal with, while the downloaded bundle is only 1.1GB. That's
    way more space overhead than I expected. I'd rather read the raw bundle
    at a different level, even if that means it won't work when using
    mercurial libraries (it will only work with the wire experiment
    enabled).
Commits on Mar 13, 2019
  1. Fix warnings when building with clang on mac

    glandium committed Mar 13, 2019
  2. [store-changegroup] Read the whole bundle first

    glandium committed Mar 13, 2019
    This is a first step in the direction of making the download as fast as
    possible while keeping the import in parallel. This is obviously not
    what happens here, as we're explicitly delaying the import after the
    entire download, making the entire affair slower, but this lays the
    ground for the parallelization work.
    
    Also, hide progress from the python code. The C code will have to report
    its own progress, but we'll implement that later.
Commits on Mar 12, 2019
  1. Move manifest check while storing to helper

    glandium committed Mar 11, 2019
Commits on Mar 11, 2019
  1. Experimentally send a changegroup stream to the helper

    glandium committed Mar 10, 2019
  2. [CI] Add more test configurations with graft enabled

    glandium committed Mar 11, 2019
Commits on Mar 8, 2019
  1. Revert "Increase the default pack depth to 50"

    glandium committed Mar 8, 2019
    This reverts commit 7bc8f5d.
    
    This was applied to git-core and is not necessary anymore. It also means
    that pack.depth will now have an effect.
Commits on Mar 7, 2019
  1. Record all helper commands to appear in fast_import_crash log

    glandium committed Mar 7, 2019
Commits on Mar 6, 2019
  1. Report the helper CMD_VERSION on the version command output.

    glandium committed Mar 6, 2019
    This should be backwards compatible with how older versions of the
    python code read the version.
  2. [CI] Print out the cinnabar version during tests

    glandium committed Mar 6, 2019
    This also tests the --version code.
Commits on Mar 5, 2019
  1. Fix off-by-one in oid_array_remove

    glandium committed Mar 5, 2019
  2. Allow to override the clonebundle url for testing

    glandium committed Mar 5, 2019
Commits on Mar 3, 2019
  1. Add optional cpu reporting

    glandium committed Mar 3, 2019
  2. Use HTTPReader instead of url.open for hg bundles

    glandium committed Mar 3, 2019
  3. [CI] Store clones as git bundles instead of tarballs

    glandium committed Mar 3, 2019
  4. Make HTTPReader more resilient to interruptions when HTTP Range reque…

    glandium committed Mar 3, 2019
    …sts are supported
  5. Only try to recover from failed HTTP bundle GET when server reports A…

    glandium committed Mar 3, 2019
    …ccept-Ranges: bytes
  6. Support git bundles as cinnabarclone source

    glandium committed Mar 3, 2019
Commits on Mar 2, 2019
  1. Move get_clonebundle.Getter to util.HTTPReader

    glandium committed Mar 2, 2019
  2. [CI] Fail tests when cinnabarclone fails

    glandium committed Mar 2, 2019
    instead of falling back (the expected behavior in that case).
  3. Only allow cinnabarclone urls with http, https or git scheme

    glandium committed Mar 2, 2019
  4. Switch to helper-enabled store-manifest

    glandium committed Mar 2, 2019
  5. Update git-core to v2.21.0

    glandium committed Mar 2, 2019
  6. Use different types and functions for hg sha1s as much as possible.

    glandium committed Mar 2, 2019
    Because git upstream is moving to support different sizes of object ids, and on
    the long run, the hg and git sizes might differ (e.g. cloning sha1 hg with
    sha256 git), we cannot continue to use the same types, considering many
    functions attached to object ids rely on the object id size of the git
    repository.
You can’t perform that action at this time.