Add test that features are enabled

[nealrichardson]

Checklist

• Used a personal 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.

Assuming this test is being run (I couldn't find where in the build logs), it should fail, given the reports in #56 and #62.

[conda-forge-webservices]

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.

[nealrichardson]

Assuming this fails the builds as expected, we could try applying nealrichardson/arrow@e925a3f as a patch and see if that fixes things. I wasn't able to confirm this in apache/arrow CI because it turns out that the conda packages we build nightly already have the features enabled correctly (compare https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=685963&view=logs&j=4f922444-fdfe-5dcf-b824-02f86439ef14&t=b2a8456a-fb11-5506-ca32-5ccd32538dc0&l=394 here to https://dev.azure.com/ursacomputing/crossbow/_build/results?buildId=47681&view=logs&j=0da5d1d9-276d-5173-c4c4-9d4d4ed14fdb&t=d9b15392-e4ce-5e4c-0c8c-b69645229181&l=21736 on apache/arrow, where the feature flags are set correctly).

[h-vetinari]

Do we need to do something like in the upstream R Makefile to discover the deps correctly?

[nealrichardson]

Do we need to do something like in the upstream R Makefile to discover the deps correctly?

No, different dependencies. https://github.com/apache/arrow/blob/apache-arrow-11.0.0/r/configure#L250-L286 is the code in question: if ArrowOptions.cmake, which is generated in the arrow-cpp build, isn't found, we won't be able to determine which features are enabled in the C++ library, so we'll essentially assume they aren't present and won't turn any of them on in the R package.

[h-vetinari]

if ArrowOptions.cmake, which is generated in the arrow-cpp build, isn't found, we won't be able to determine which features are enabled in the C++ library, so we'll essentially assume they aren't present and won't turn any of them on in the R package.

OK cool. That file is definitely being packaged with libarrow, so it should just be a question of pointing the R build to $PREFIX/lib/cmake/Arrow/ArrowOptions.cmake?

[nealrichardson]

Right. And in theory the patch I referenced would do that.

[h-vetinari]

Right. And in theory the patch I referenced would do that.

I was hoping it might be as easy as setting that variable correctly, but alas no. Shouldn't an environment variable be picked up by the R configure script though?

[nealrichardson]

Hmm, based on the build log output, LIB_DIR is being set in the script and overwriting the env var you set, but interestingly there's no " around that line like the lines above it: https://github.com/apache/arrow/blob/main/r/configure#L113

I don't understand the subtleties of $PREFIX here, but I just see that the PKG_LIBS="${PKGCONFIG_LIBS}" seems to do what is expected but LIB_DIR=${FOUND_LIB_DIR} does not, and we know from the log message printed by L109 that FOUND_LIB_DIR is literally $PREFIX/lib https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=695884&view=logs&j=d0d954b5-f111-5dc4-4d76-03b6c9d0cf7e&t=841356e0-85bb-57d8-dbbc-852e683d1642&l=294.

[h-vetinari]

Turns out I had forgotten to add /lib to the $PREFIX, now it works! :)

I think I'm going to merge this to unblock the two issues that were reported; it should be a net improvement in all respects already, and we can still keep iterating in a new PR if necessary.

[nealrichardson]

Hmm, based on the build log output, LIB_DIR is being set in the script and overwriting the env var you set, but interestingly there's no " around that line like the lines above it: https://github.com/apache/arrow/blob/main/r/configure#L113

I don't understand the subtleties of $PREFIX here, but I just see that the PKG_LIBS="${PKGCONFIG_LIBS}" seems to do what is expected but LIB_DIR=${FOUND_LIB_DIR} does not, and we know from the log message printed by L109 that FOUND_LIB_DIR is literally $PREFIX/lib https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=695884&view=logs&j=d0d954b5-f111-5dc4-4d76-03b6c9d0cf7e&t=841356e0-85bb-57d8-dbbc-852e683d1642&l=294.

For posterity, it turns out that I was looking at the latest main branch, where this issue had been fixed: apache/arrow#34229

So the workaround added here can be removed in the 12.0.0 release (likely happening within days). Having it here won't harm anything as it will get overwritten in the configure script, as I described. And the test we added here will confirm that it is safe to delete the workaround.