Blacklist pillow 7.1.*

[phue]

There is a regression in the latest Pillow releases

python-pillow/Pillow#4528
The fix has been merged but not released yet

Checklist

• Used a fork of the feedstock to propose changes
• Bumped the build number (if the version is unchanged)
• Reset the build number to 0 (if the version changed)
• Re-rendered with the latest conda-smithy (Use the phrase @conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
• Ensured the license file is being packaged.

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

I do have some suggestions for making it better though...

For recipe:

• License is not an SPDX identifier (or a custom LicenseRef) nor an SPDX license expression.

[phue]

@conda-forge-admin, please rerender

[jni]

Thanks for handling this, @phue!

I must say I don't understand the CI failures... Two appear to be some numerical precision issues that for some reason don't appear in the other builds, and the last one appears to be a dependency conflict.

The numerical precision issues could be easily dealt with with a patch, but I don't know how to deal with the dependency conflicts. @jakirkham? 😬

[phue]

@conda-forge-admin, please rerender

[phue]

Thanks for handling this, @phue!

I must say I don't understand the CI failures... Two appear to be some numerical precision issues that for some reason don't appear in the other builds, and the last one appears to be a dependency conflict.

The numerical precision issues could be easily dealt with with a patch, but I don't know how to deal with the dependency conflicts. @jakirkham? grimacing

I don't understand them either.
Rerendering with latest conda-smithy made them disappear again. But now there is another error which I also don't understand, this time related to tcl

[hmaarrfk]

I'm not sure if this is going to have the effect you want.

THese kinds of issues are rather typical in rolling releases like conda-forge.

Maybe you can ask the package to be removed if the issue is serious enough?

[jakirkham]

I'm think we should just do a repodata patch if the issue is this version of pillow has issues. We could then apply this to both the current scikit-image package and older versions. WDYT?

[hmaarrfk]

repodata patch

I'm not sure what this is.

[jakirkham]

The idea would be to patch repodata for the conda-forge channel to alter the requirements of scikit-image packages. Patching is applied based on code here. Here's a somewhat recent patch that distributed needed as an example ( conda-forge/conda-forge-repodata-patches-feedstock#32 ).

[hmaarrfk]

that sounds harder than simply pulling pillow 7.1

[jakirkham]

What about applying the upstream fix as a patch to pillow then?

[hmaarrfk]

What about applying the upstream fix as a patch to pillow then?

I'm not super in the patching mood today, but yeah, that would probably be easiest.

I also don't have power on the Pillow feedstock, so :/

[phue]

I'm not sure if this is going to have the effect you want.

THese kinds of issues are rather typical in rolling releases like conda-forge.

Could you elaborate more on this? I currently don't understand why blacklisting is not a viable option. And I don't see how the CI failure is related to the change I made

[hmaarrfk]

I currently don't understand why blacklisting is not a viable option.

The solver has been changing recently, so if somebody types conda install pillow then types conda install scikit-image I think they will get pillow 7.1.

And I don't see how the CI failure is related to the change I made

Neither do I. I just skipped the test.

What platform do you use? I could just merge this even if it is failing if you want....

[hmaarrfk]

honestly, i'm not too sure how the solver works out, so i'm willing to try this out, I just wanted to say that typically, the best way is to ping upstream for a fix, or to patch up the offending repo.

[phue]

The solver has been changing recently, so if somebody types conda install pillow then types conda install scikit-image I think they will get pillow 7.1.

OK, now I get it that's unfortunate. I think the current situation is even worse though, because anyone typing conda install scikit-image will get a dysfunctional installation

honestly, i'm not too sure how the solver works out, so i'm willing to try this out, I just wanted to say that typically, the best way is to ping upstream for a fix, or to patch up the offending repo.

I fully agree on that, in this case the fix has already been merged upstream so it's just pending a new release. Patching the pillow feedstock is of course an option, I just thought this would be in line with what was done for pip here

[hmaarrfk]

it seems that you are hitting a nasty bug on OSX.. even if i merged, it would give you inconstient result on different platforms

[hmaarrfk]

i'm ok skipping those tk tests, but not so sure about skipping the OSX test.

I'll let @jni review before merging

[phue]

Interesting, but this will affect the next scikit-image release as well, regardless of the pillow breakage.

[hmaarrfk]

i totally agree, but that is a "tomorrow" problem ;)

[hmaarrfk]

Pillow 7.1.2 has just been released.

I'm inclined to close this for today. What do you think?

[phue]

Absolutely, good that it is now resolved upstream

[hmaarrfk]

Nice. Yeah. Itis really nice when projects can respond to quick bugs like that