c-blosc upgrade 1.18.1 -> 1.21.0

[olly-writes-code]

c-blosc submodule upgrade 1.18.1 -> 1.21.0 to fix issue #269

TODO:

• Unit tests and/or doctests in docstrings
• tox -e py39 passes locally
• Docstrings and API docs for any new/modified user-facing classes and functions
• Changes documented in docs/release.rst
• tox -e docs passes locally
• GitHub Actions CI passes
• Test coverage to 100% (Coveralls passes)

[olly-writes-code]

When trying to run on my local machine I got the following error

      c-blosc/blosc/blosc.c:31:12: fatal error: 'snappy-c.h' file not found
        #include "snappy-c.h"
                 ^~~~~~~~~~~~ `

[olly-writes-code]

This fixes the file not found snappy issue when compiling c-blosc 6df653b

IDK - how this will play with other machines.

[jakirkham]

Yeah I don't think we want to skip using Snappy. Did this move somewhere else? Is there a path we need to update?

[jakirkham]

Hmm...looks like they dropped Snappy

Blosc/c-blosc#295

[jakirkham]

Suggested change

	# ('HAVE_SNAPPY', 1),
	('HAVE_SNAPPY', 0), # Blosc 1.19.0+ dropped Snappy

[olly-writes-code]

I'll make this change and see if installation still succeeds

[olly-writes-code]

Confirming that it did not work with the above change #283 (comment) . I get the same error #283 (comment)

[jakirkham]

Just to confirm, that's with 0 and not 1, correct? If so, that's really weird

Could you please try None as well?

[olly-writes-code]

yep set to 0. Yeah I can try None too.

[olly-writes-code]

Yep same issue with None. I can dig into how the macro passes information to the c-blosc build process

[olly-writes-code]

Oh it seems like @joshmoore started on this very same PR but stopped because of a call #259

Do we want to proceed with this change or go down another path?

[joshmoore]

@kindjacket: explaining the #259 / #274 decisions -- we discussed that on that call that spending our time on the pure python implementation was probably a bigger win. However, as you can see in my #274, that turned out to be more work than (I) expected.

At this point, I'm inclined to say we temporarily drop snappy in order to get M1 fixed and then re-evaluate once that's working with the options being continue on the #274 path, or to first work on getting snappy working with c-block 1.21+.

@zarr-developers/core-devs : any objections?

[joshmoore]

@kindjacket : looks like we have a 👍 from @jakirkham as well (#268), any thoughts on getting this ready to merge?

[pep8speaks]

Hello @kindjacket! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2021-08-18 09:31:52 UTC

[joshmoore]

whew Starting to look greenish.

Edit: CondaHTTPError: HTTP 502 BAD GATEWAY for url <https://conda.anaconda.org/conda-forge/noarch/current_repodata.json> looks like those that aren't may be unrelated. I'm trying to restart them into greenness.

[jakirkham]

Actually can we leave the fixture data (instead of deleting it)? We will want this when we add back snappy (somehow 🤔)

[joshmoore]

Actually can we leave the fixture data (instead of deleting it)? We will want this when we add back snappy

Good point. I didn't realize at first that it was an ordered list etc. etc. so I thought deletion was the appropriate way forward. I'll resurrect.

[joshmoore]

Any thoughts on version number with the drop of snappy? 0.9.0 I assume.

[joshmoore]

I'm taking that at least as no objections.

[jakirkham]

Thanks Josh! 😄

Sorry had to drop 😴 Though happy to trust your judgement on versions 🙂

Raised issue ( #285 ) about readding Snappy

[joshmoore]

np.

A heads up that while testing zarr-developers/zarr-python#773 I'm pretty sure that this PR changed the checksum of chunks in the zarr_implementations test.