Rebuild for hdf5 1.14.4

[regro-cf-autotick-bot]

This PR has been triggered in an effort to update hdf51144.

Notes and instructions for merging this PR:

1. Please merge the PR only after the tests have passed.
2. Feel free to push to the bot's branch to update this PR if needed.

Please note that if you close this PR we presume that the feedstock has been rebuilt, so if you are going to perform the rebuild yourself don't close this PR until the your rebuild has been merged.

---

If this PR was opened in error or needs to be updated please add the bot-rerun label to this PR. The bot will close this PR and schedule another one. If you do not have permissions to add this label, you can use the phrase @conda-forge-admin, please rerun bot in a PR comment to have the conda-forge-admin add it for you.

_{This PR was created by the regro-cf-autotick-bot. The regro-cf-autotick-bot is a service to automatically track the dependency graph, migrate packages, and propose package version updates for conda-forge. Feel free to drop us a line if there are any issues! This PR was generated by https://github.com/regro/cf-scripts/actions/runs/11171199416 - please use this URL for debugging.}

[conda-forge-admin]

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/meta.yaml) and found it was in an excellent condition.

[gillins]

Hi @petebunting any idea why this would fail on OSX? Have you had any trouble with hdf5 1.14.4 there?

[petebunting]

Hi @gillins, No is the simple answer. I've not had any problems with rsgislib building against hdf5. I only have ARM Macs these days, but I can try a local build later today to see if I can replicate it.

[gillins]

It's weird because osx_arm64 succeeded...

[gillins]

Ah actually osx arm64 is cross compiled so it doesn't run the tests. Thanks for having a look Pete - interested to hear what you find with hdf5 1.14.4...

[petebunting]

@gillins it didn't look to me that the test was running on OSX arm where the error was occurring... If I remember correctly from setting up the rsgislib osx arm build it is difficult to run the tests on osx arm as it cross compiled or something. If it isn't running the test then it might not be finding the error...

[gillins]

@petebunting just seen this: conda-forge/hdf5-feedstock#232 - might be the problem? Weird how it was only on OSX for us...

[gillins]

@conda-forge-admin, please restart ci​

[gillins]

@jschueller any ideas about this crash with HDF5 C++? Seems different to the one you've recently fixed...

[jschueller]

dont know, maybe its due to clang17 since its only happening on osx

[xylar]

I had to roll back to clang 15 elsewhere (conda-forge/hdfeos5-feedstock#42 and conda-forge/libcdms-feedstock#56). Maybe that's the way forward here, too, so the migration can progress?

[gillins]

@conda-forge-admin, please rerender

[xylar]

Okay, not that then. Sorry for leading you astray.

[gillins]

I'm now getting compile errors, rather than segfaults on osx:

/Users/runner/miniforge3/conda-bld/kealib-split_1728351685884/work/include/libkea/KEAAttributeTable.h:34:10: fatal error: 'iostream' file not found
#include <iostream>

Any thoughts?

[xylar]

I'm afraid I don't have any idea. I'd suggest going back to clang 17. I now see that worked fine for you in #64 so no reason to go back to an earlier version.

I would suggest trying to get more verbose output from ctest instead (after going back to clang 17).

[gillins]

@conda-forge-admin, please rerender

[xylar]

Unfortunately, not hugely informative:

1/1 Test #1: test1 ............................Subprocess aborted***Exception:   0.03 sec
libc++abi: terminating with uncaught exception of type H5::FileIException

[gillins]

It's so weird that it's only osx... @petebunting if you could get a chance to look at this and see what is going on that would be great, otherwise I guess we'll have to skip osx (or maybe just the tests but that seems dangerous) for now so the migration can continue.
Thanks for your help @xylar @jschueller

[gillins]

@conda-forge-admin, please rerender

[gillins]

@conda-forge-admin, please rerender

[gillins]

@conda-forge-admin, please rerender

[gillins]

@conda-forge-admin, please rerender

[gillins]

@h-vetinari is it possible because of the way I'm running ctest that the wrong libc++abi is being picked up somehow (or a mix)? Should I be setting DYLD_LIBRARY_PATH or similar?

[akrherz]

I stumbled into this PR being important for conda-forge dependency troubles attm, so hopefully you don't mind a drive-by rerender to see if this somehow works now :/

@conda-forge-admin, please rerender

[akrherz]

Looking at the failed log, this combination of packages seems strange to me, installed with the same buildstep

libcxx:                          19.1.2-hf95d169_0         conda-forge
libcxx-devel:                    17.0.6-h8f8a49f_6         conda-forge

I am likely just making noise, but try adding the libcxx-devel dep to see it is straightens this out?

[petebunting]

Hi @gillins Sorry it has taken me such a long while to look at this again. As you say, I think it is a compiler issue. I've just done a local build of both hdf 1.14.5 (latest from hdf5 releases) and kea 1.6.0 and run the kea test1 binary and it worked returning 'Success', which it wasn't doing using the conda-forge hdf5 binary so I don't think it is a bug or something within the hdf5 code.

Tested on an M2 MBP.

[xylar]

@petebunting, I had suspected it was something to do with how hdf5 is being built on conda-forge, but we did tests with both 1.14.4 and 1.14.5 and were not able to reproduce the issue over there:
conda-forge/hdf5-feedstock#234
conda-forge/hdf5-feedstock#233

I also re-tested with 1.14.3 here and didn't see the error so it doesn't seem to be something that crept into the compilers since your last kealib build:
#66

So I'm afraid I haven't been able to narrow down where it's coming from.

[gillins]

@petebunting can you tell if more than one copy of libc++ is being loaded when you run the failing conda forge tests?

I guess we are going to have to decide if we skip osx and allow the migration to proceed, or we wait until more information if available...

[xylar]

I guess we are going to have to decide if we skip osx and allow the migration to proceed, or we wait until more information if available...

I do not recommend this approach, as it means skipping OSX in all downstream migrations. We ended up in a nightmare but doing something similar with ppc64le in the recent past (in that case for the proj library) and the trouble was more than we faced in the end by just taking the time to figure out the problem. It is better to have the migration stuck than to have it proceed only half working.

[h-vetinari]

@h-vetinari is it possible because of the way I'm running ctest that the wrong libc++abi is being picked up somehow (or a mix)? Should I be setting DYLD_LIBRARY_PATH or similar?

Very sorry that I missed this ping. I see you have Isuru's attention already, so you should be on great hands. :)