-
Notifications
You must be signed in to change notification settings - Fork 580
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
CI: Test pre-release versions of dependencies #2224
Conversation
140ffcb
to
b6acc3c
Compare
Absolutely, this is really useful. I wanted to suggest this on the the nibabel thread. Maybe we should even consider testing nibabel master somewhere? I just wonder if it should live in nibabel or nilearn? |
+1 for merge once CI has passed. |
I think it's usually easier for downstream projects to test pre-releases of upstream, since you already have all the tools in place to test nilearn, and can also catch other tools' pre-releases. Speaking of which, numpy/scipy and a number of other projects also push nightly wheels to http://8b8c47f30575e674b56d-47bd50c35cd79bd838daf386af554a83.r59.cf2.rackcdn.com/, if you want to test against even more bleeding-edge changes. |
I think it's usually easier for downstream projects to test pre-releases of upstream, since you already have all the tools in place to test nilearn, and can also catch other tools' pre-releases.
But then changes in nibabel break _our_ CI, which I am not comfortable with, as in my view these changes are bugs in nibabel.
Speaking of which, numpy/scipy and a number of other projects also push nightly wheels
We haven't really had difficulties with changes in these tools. Maybe scikit-learn, to be fair (which is ironical, as I am also core dev of scikit-learn)
|
I had opened an issue for this and never implemented it. Thanks Chris!
…On Wed, Nov 20, 2019, 20:45 Gael Varoquaux ***@***.***> wrote:
> I think it's usually easier for downstream projects to test pre-releases
of upstream, since you already have all the tools in place to test nilearn,
and can also catch other tools' pre-releases.
But then changes in nibabel break _our_ CI, which I am not comfortable
with, as in my view these changes are bugs in nibabel.
> Speaking of which, numpy/scipy and a number of other projects also push
nightly wheels
We haven't really had difficulties with changes in these tools. Maybe
scikit-learn, to be fair (which is ironical, as I am also core dev of
scikit-learn)
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#2224?email_source=notifications&email_token=AB6SXRCGBDJN54R7TQMH7D3QUWHWPA5CNFSM4JPYDAS2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEEUIDTY#issuecomment-556302799>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AB6SXRBAAXBWXX5BTIOQJRTQUWHWPANCNFSM4JPYDASQ>
.
|
Codecov Report
@@ Coverage Diff @@
## master #2224 +/- ##
=========================================
Coverage ? 92.35%
=========================================
Files ? 150
Lines ? 19004
Branches ? 2309
=========================================
Hits ? 17552
Misses ? 932
Partials ? 520 Continue to review full report at Codecov.
|
@effigies Thanks for this!
Should this be "allowed to fail"? Azure Pipelines(AZP) has spare containers, moving this there would be better. If you don't wanna mess with AZP settings I can do it after this one is merged. |
Marked as an allowed failure. I'm definitely not feeling up to working on an Azure test right now. |
8171731
to
a7f33a2
Compare
…ix_package_finder * 'master' of https://github.com/nilearn/nilearn: CI: Test pre-release versions of dependencies (nilearn#2224) Fix: nilearn.image.smooth_img fails if fwhm is an ndarray (nilearn#2107) Update Python version warnings for Py35 deprecation (nilearn#2214) Fixing failure in minimum build of TravisCI (nilearn#2227) Updated missing whats_new entry
…s-craddock * 'master' of https://github.com/nilearn/nilearn: (620 commits) CI: Test pre-release versions of dependencies (nilearn#2224) Fix: nilearn.image.smooth_img fails if fwhm is an ndarray (nilearn#2107) Update Python version warnings for Py35 deprecation (nilearn#2214) Fixing failure in minimum build of TravisCI (nilearn#2227) Updated missing whats_new entry Add testing for Nilearn setup & install & Fix the broken dependency installation (nilearn#2201) Fix uniform ball cloud test for sklearn >= 0.22 (nilearn#2175) MAINT: sklearn Deprecations (nilearn#2219) DOC: title in bold [DOC] Add note about decreasing memory usage (nilearn#2223) Rel 060b0 (nilearn#2208) Nilearn 0.6.0b0 release (nilearn#2206) Fixed the redundant & missing test case in merged PR nilearn#2035 (nilearn#2205) Modify fetch_development_fmri to fetch adults or children (nilearn#2035) Verbose doc building to ease tracking of progress & diagnose stalls (nilearn#2203) New conda env is created once conda path has been activated Conda environment is created for full-builds Refactor CircleCI config for reduced redundancy (nilearn#2204) Installation should fail on Python < 3.5 (nilearn#2198) [MRG] Add get data function (nilearn#2172) ... # Conflicts: # nilearn/datasets/atlas.py
…niftis * 'master' of https://github.com/nilearn/nilearn: Argument version='det' in pauli 2017 atlas now works correctly (nilearn#2235) Fix: Nilearn raises Nose ImportError with no MatPlotLib (nilearn#2231) Non-iterable Image headers are not copied anymore (nilearn#2212) Add parameter "strategy" to NiftiLabelsMasker to choose function to reduce ROIs (nilearn#2221) Improved nifti masker documentation (nilearn#2234) CI: Use pkg_resources to list installed dependencies (nilearn#2228) CI: Test pre-release versions of dependencies (nilearn#2224) # Conflicts: # nilearn/regions/signal_extraction.py
…der2 * 'master' of https://github.com/nilearn/nilearn: BUG: NiftiLabelsMasker inverse_transform without transform (nilearn#2248) Details in CONTRIBUTING.rst Update CONTRIBUTING.rst (nilearn#2262) [MRG+1] Remove support for old parameters in view_connectome & view_markers (nilearn#2255) maint: warn upcoming deprecation for fetch_haxby_simple (nilearn#2256) fix: remove now duplicated report scraper (nilearn#2261) Region means are no longer casted to Int type for Int images in NiftLabelsMasker (nilearn#2195) Improve replace parameters (nilearn#2254) Fix tests pertaining to replace_parameters functionality (nilearn#2253) [doc] Warning for non-matching n_parcels (nilearn#2240) More clear CONTRIBUTING.rst (nilearn#2246) doc: remove now unecessary print statements (nilearn#2243) [doc] Iimprove documentation re: using index_img with slice() (nilearn#2189) Argument version='det' in pauli 2017 atlas now works correctly (nilearn#2235) Fix: Nilearn raises Nose ImportError with no MatPlotLib (nilearn#2231) Non-iterable Image headers are not copied anymore (nilearn#2212) Add parameter "strategy" to NiftiLabelsMasker to choose function to reduce ROIs (nilearn#2221) Improved nifti masker documentation (nilearn#2234) CI: Use pkg_resources to list installed dependencies (nilearn#2228) CI: Test pre-release versions of dependencies (nilearn#2224)
Fixes #1728
I noticed that you aren't testing pre-releases (I was checking to see if the nibabel RC was causing problems). I've found it very helpful to see when changes coming down the pipeline (particularly from numpy) are going to break compatibility.
This PR adds an optional
PIP_FLAGS
environment variable that will be added topip install
commands for requirements and nilearn.Feel free to merge if this seems useful. I can also just use the Travis results for this PR to see if we're breaking anything.