search

conda-forge / conda-forge.github.io

The conda-forge website.
https://conda-forge.org
BSD 3-Clause "New" or "Revised" License
132 stars 280 forks source link

Clean up feedstocks still using `conda-forge-ci-setup=3` #2282

Open h-vetinari opened 3 months ago

h-vetinari commented 3 months ago

I noticed this in pytorch, and according to a search for this, we have around 10 feedstocks that are currently affected.

The problem with this is that using such an old ci-setup forces a very old conda-build, which is - pretty much by sheer luck - just barely new enough to correctly handle {{ stdlib("c") }}, which we need to avoid creating packages with incorrect metadata on osx/linux.

There are many other relevant changes that should make us retry to make these feedstocks use up-to-date build tools (some may still need to keep pinning lief due to various issues, but aside from a performance regression on osx, lief 0.14 should have fixed most things).

Krande commented 3 months ago

@h-vetinari Okay, so all we need to do is to remove the remote_ci_setup entries and check to see if all goes well?

We are currently preparing a new release of ifcopenshell in https://github.com/conda-forge/ifcopenshell-feedstock/pull/54. So I will try to make the changes there to see if that goes well.

Here's what we have. I remember pinning py-lief was necessary way back when due to the osx builds ran out of time. Changing the py-lief version at the time fixed it. Hopefully this is gone now.

Here's the remote_ci_setup in ifcopenshell

remote_ci_setup:
- conda-forge-ci-setup=3
- py-lief>=0.12.3
Tobias-Fischer commented 3 months ago

pcl, orocos-kdl, libignition-common all done

Krande commented 3 months ago

@h-vetinari ifcopenshell is updated with the outdated configs removed.

bollwyvl commented 3 months ago

tlsh is fixed (well, building)... I could probably use an assist on https://github.com/conda-forge/atheris-feedstock/pull/16 (my c++ chops are non-existent):

src/native/counters.cc:26:6: error: variable or field '__sanitizer_cov_8bit_counters_init' declared void
   26 | void __sanitizer_cov_8bit_counters_init(uint8_t* start, uint8_t* stop);
      |      ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
src/native/counters.cc:26:41: error: 'uint8_t' was not declared in this scope
   26 | void __sanitizer_cov_8bit_counters_init(uint8_t* start, uint8_t* stop);
      |                                         ^~~~~~~
src/native/counters.cc:24:1: note: 'uint8_t' is defined in header '<cstdint>'; did you forget to '#include <cstdint>'?
   23 | #include "macros.h"
  +++ |+#include <cstdint>
Tobias-Fischer commented 3 months ago

@bollwyvl: this patch should fix it: https://patch-diff.githubusercontent.com/raw/google/atheris/pull/91.patch

bollwyvl commented 3 months ago

Thanks! Still seeing some OSX issues on atheris (but not missing libfuzzer, which was happening previously)