py-h5py: add 3.8.0 by michaelkuhn · Pull Request #35960 · spack/spack

michaelkuhn · 2023-03-09T10:08:16Z

No description provided.

tldahlgren

Confirmed the sha256.

adamjstewart · 2023-03-10T01:56:11Z

Also being updated in #35195

michaelkuhn · 2023-03-10T14:44:02Z

Also being updated in #35195

Should we update it separately or wait for #35195 (which seems to take a while)?

michaelkuhn · 2023-03-10T14:44:07Z

@spackbot run pipeline

spackbot-app · 2023-03-10T14:44:15Z

I've started that pipeline for you!

adamjstewart · 2023-03-10T15:28:11Z

I'm fine with updating it separately but this PR is missing some of the changes in the other PR (upper bounds on dep versions, etc.). If you want to steal them from that PR go ahead.

michaelkuhn · 2023-03-10T15:41:44Z

I've pushed a new version that should now also include the correct upper bounds. I've left the setuptools@61: dependency commented out since that seems to cause problems for @kwryankrattiger.

kwryankrattiger · 2023-03-10T15:46:11Z

I am okay with it being separate, I wasn't too thrilled about updating it in #35195 since it is outside of the scope of the changes in there.

var/spack/repos/builtin/packages/py-h5py/package.py

takluyver · 2023-03-12T07:21:18Z

The pinning is kind of a workaround - we aim to ensure it builds with an older version, and then it will be compatible with newer versions at runtime. The pinned versions are roughly the first mpi4py versions that will conveniently work for each python version. I imagine that Spack has better ways to express this stuff. Needless to say, our own package metadata is a compromise between a pure description and making things work for python tools which use that metadata directly.

…

On Sat, 11 Mar 2023, 21:31 Adam J. Stewart, ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In var/spack/repos/builtin/packages/py-h5py/package.py <#35960 (comment)>: > depends_on("py-six", type=("build", "run"), when="@:2") # Link dependencies (py-h5py v2 cannot build against HDF5 1.12 regardless # of API setting) ***@***.***:1.11 +hl", when="@:2") - ***@***.***:1.12 +hl", ***@***.***:") + ***@***.***:1.12 +hl", ***@***.***:3.7") + ***@***.***:1.14 +hl", ***@***.***:") # MPI dependencies depends_on("hdf5+mpi", when="+mpi") mpi4py dep needs to be updated too: RUN_REQUIRES.append('mpi4py >=3.0.2') SETUP_REQUIRES.append("mpi4py ==3.0.2; python_version<'3.8'") SETUP_REQUIRES.append("mpi4py ==3.0.3; python_version=='3.8.*'") SETUP_REQUIRES.append("mpi4py ==3.1.0; python_version=='3.9.*' or python_version=='3.10.*'") SETUP_REQUIRES.append("mpi4py ==3.1.4; python_version>='3.11'") Honestly, this stuff is kinda annoying to track. Not sure why things need to be pinned to a specific version for each Python release. Should we just ignore this and say ***@***.***: for all versions? Curious what @bryanherman <https://github.com/bryanherman> @takluyver <https://github.com/takluyver> think. — Reply to this email directly, view it on GitHub <#35960 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACQB5J3HTN2D3QVQJILOFDW3TVKXANCNFSM6AAAAAAVU4MAZI> . You are receiving this because you were mentioned.Message ID: ***@***.***>

adamjstewart · 2023-03-12T17:12:30Z

In that case I'll let @michaelkuhn decide what to do. It sounds like [email protected]: should be sufficient. We don't currently use any pip/build flags to ensure that dependencies are correct, but we might in the future. If we do, I can always change this to be technically correct later.

The pinned versions are roughly the first mpi4py versions that will conveniently work for each python version.

I'm guessing this means the first version where wheels are available, not the first version that supports that Python version? If it was the latter I would add these to the py-mpi4py package so we can remove old versions when we deprecate Python versions.

takluyver · 2023-03-20T11:52:56Z

I'm guessing this means the first version where wheels are available, not the first version that supports that Python version?

That would be my guess too, though I don't know that we've applied that rule completely consistently.

michaelkuhn · 2023-03-28T19:03:17Z

Sorry for dropping the ball on this. I've just pushed a new version that constrains mpi4py to @3.0.2: for all 3.x versions.

var/spack/repos/builtin/packages/py-h5py/package.py

adamjstewart

Looks fine to me as long as the other 2 maintainers don't see any issues.

takluyver · 2023-03-29T08:30:03Z

LGTM.

The precisely versioned dependencies on numpy and Cython could probably be simplified, like with mpi4py (discussed above). They're similar cases where we've found compromises between a pure theoretical description of dependencies, and what practically works for Python-native tooling. But if they're not causing an issue, you may just want to leave them as they are and not risk breaking anything.

kwryankrattiger · 2023-03-29T15:54:57Z

@spackbot run pipeline

spackbot-app · 2023-03-29T15:55:03Z

I've started that pipeline for you!

kwryankrattiger · 2023-03-30T19:53:25Z

if no objections in the next couple hours, I am going to merge this today.

spackbot-app bot added dependencies new-version python update-package labels Mar 9, 2023

spackbot-app bot requested review from adamjstewart, bryanherman, pradyunsg and takluyver March 9, 2023 10:08

takluyver previously approved these changes Mar 9, 2023

View reviewed changes

tldahlgren reviewed Mar 9, 2023

View reviewed changes

tldahlgren assigned takluyver and tldahlgren Mar 9, 2023

adamjstewart self-assigned this Mar 10, 2023

michaelkuhn dismissed takluyver’s stale review via a001ced March 10, 2023 15:40

michaelkuhn force-pushed the py-h5py-3.8.0 branch from f95e61c to a001ced Compare March 10, 2023 15:40

michaelkuhn force-pushed the py-h5py-3.8.0 branch from a001ced to 7b7a46e Compare March 10, 2023 15:43

adamjstewart reviewed Mar 10, 2023

View reviewed changes

var/spack/repos/builtin/packages/py-h5py/package.py Outdated Show resolved Hide resolved

michaelkuhn force-pushed the py-h5py-3.8.0 branch from 7b7a46e to 4e641f1 Compare March 10, 2023 21:34

adamjstewart reviewed Mar 11, 2023

View reviewed changes

var/spack/repos/builtin/packages/py-h5py/package.py Outdated Show resolved Hide resolved

var/spack/repos/builtin/packages/py-h5py/package.py Show resolved Hide resolved

michaelkuhn force-pushed the py-h5py-3.8.0 branch 2 times, most recently from 4cd1d12 to 40c13ec Compare March 11, 2023 18:53

adamjstewart reviewed Mar 11, 2023

View reviewed changes

var/spack/repos/builtin/packages/py-h5py/package.py Show resolved Hide resolved

kwryankrattiger mentioned this pull request Mar 27, 2023

Ecp hdf5 vol #35195

Merged

michaelkuhn force-pushed the py-h5py-3.8.0 branch from 40c13ec to b6fe825 Compare March 28, 2023 19:02

adamjstewart reviewed Mar 28, 2023

View reviewed changes

var/spack/repos/builtin/packages/py-h5py/package.py Outdated Show resolved Hide resolved

py-h5py: add 3.8.0

1cdec63

michaelkuhn force-pushed the py-h5py-3.8.0 branch from b6fe825 to 1cdec63 Compare March 28, 2023 21:19

adamjstewart approved these changes Mar 28, 2023

View reviewed changes

adamjstewart merged commit 996442e into spack:develop Mar 31, 2023

Conversation

michaelkuhn commented Mar 9, 2023

Uh oh!

tldahlgren left a comment

Choose a reason for hiding this comment

Uh oh!

adamjstewart commented Mar 10, 2023

Uh oh!

michaelkuhn commented Mar 10, 2023

Uh oh!

michaelkuhn commented Mar 10, 2023

Uh oh!

spackbot-app bot commented Mar 10, 2023

Uh oh!

adamjstewart commented Mar 10, 2023

Uh oh!

michaelkuhn commented Mar 10, 2023

Uh oh!

kwryankrattiger commented Mar 10, 2023

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

takluyver commented Mar 12, 2023 via email

Uh oh!

adamjstewart commented Mar 12, 2023

Uh oh!

takluyver commented Mar 20, 2023

Uh oh!

michaelkuhn commented Mar 28, 2023

Uh oh!

Uh oh!

adamjstewart left a comment

Choose a reason for hiding this comment

Uh oh!

takluyver commented Mar 29, 2023

Uh oh!

kwryankrattiger commented Mar 29, 2023

Uh oh!

spackbot-app bot commented Mar 29, 2023

Uh oh!

kwryankrattiger commented Mar 30, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants