diff options
author | Richard van der Hoff <richard@matrix.org> | 2018-10-25 01:00:58 +0100 |
---|---|---|
committer | Richard van der Hoff <richard@matrix.org> | 2018-10-25 01:00:58 +0100 |
commit | 46f98a6a297ff014e7e88fb056e59755fd18cf58 (patch) | |
tree | 928d39b9d266b9ca6611b2d4e5e2a1baebc8ad9f /.travis.yml | |
parent | Combine the pep8 and check_isort builds into one (diff) | |
download | synapse-46f98a6a297ff014e7e88fb056e59755fd18cf58.tar.xz |
Only cache the wheels
Diffstat (limited to '.travis.yml')
-rw-r--r-- | .travis.yml | 12 |
1 files changed, 10 insertions, 2 deletions
diff --git a/.travis.yml b/.travis.yml index 64079b56f6..ac34c36725 100644 --- a/.travis.yml +++ b/.travis.yml @@ -1,8 +1,16 @@ sudo: false language: python -# tell travis to cache ~/.cache/pip -cache: pip +cache: + directories: + # we only bother to cache the wheels; parts of the http cache get + # invalidated every build (because they get served with a max-age of 600 + # seconds), which means that we end up re-uploading the whole cache for + # every build, which is time-consuming In any case, it's not obvious that + # downloading the cache from S3 would be much faster than downloading the + # originals from pypi. + # + - $HOME/.cache/pip/wheels # don't clone the whole repo history, one commit will do git: |