Try to trigger new boost build #123

jan-janssen · 2020-12-22T09:20:27Z

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.

conda-forge-linter · 2020-12-22T09:20:33Z

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.

Documentation on acceptable licenses can be found here.

jan-janssen · 2020-12-22T09:20:35Z

@conda-forge-admin, please rerender

…da-forge-pinning 2020.12.21.13.42.08

jan-janssen · 2020-12-22T09:28:19Z

@conda-forge-admin, please rerender

github-actions · 2020-12-22T09:30:08Z

Hi! This is the friendly automated conda-forge-webservice.
I tried to rerender for you, but it looks like there was nothing to do.

…da-forge-pinning 2020.12.21.13.42.08

jan-janssen · 2020-12-22T09:58:13Z

@conda-forge-admin, please rerender

github-actions · 2020-12-22T10:00:24Z

Hi! This is the friendly automated conda-forge-webservice.
I tried to rerender for you, but it looks like there was nothing to do.

mbargull · 2020-12-22T12:11:52Z

@conda-forge-admin, please rerender

mbargull · 2020-12-22T12:13:29Z

recipe/meta.yaml

@@ -164,8 +164,7 @@ outputs:
        - suitesparse
        - zlib
        - fenics-ffc =={{ version }}
-        - hdf5 * {{ mpi_prefix }}_*
-        - {{ pin_compatible("boost-cpp", exact=True) }}


{{ pin_compatible(..., exact=True) }} means that it pins to the exact build (i.e., on the build string level).
Here you rather want a plain boost-cpp in host/run so that conda-build will replace it with the pinnings definition.

The fenics package requires the exact boost-cpp version which is the cause of conda-forge/mshr-feedstock#18

exact boost-cpp version

Exact version would be understandable. But the current conda-forge-pinning already has a max_pin: x.x.x for boost-cpp which means if you have it in host and run, then you'd get a pin to >=1.72.0,<1.73.0a0.
Exact build is what doesn't seem sensible to me.

If you don't trust conda-forge-pinning (e.g., to avoid a max_pin change affecting this recipe), you can do {{ pin_compatible("boost-cpp", max_pin="x.x.x") }} yourself.

…da-forge-pinning 2020.12.22.11.23.16

mbargull · 2020-12-22T12:14:07Z

recipe/meta.yaml

@@ -164,8 +164,7 @@ outputs:
        - suitesparse
        - zlib
        - fenics-ffc =={{ version }}
-        - hdf5 * {{ mpi_prefix }}_*


hdf5 has a run_export has includes the MPI variant. We can rely on that one.

mbargull · 2020-12-22T12:16:40Z

@jan-janssen, I don't know why conda-smithy splits the build jobs for osx for each boost-cpp. But the linux build also creates two (boost-cpp-dependent) builds for fenics-libdolfin. Any downstream issue you encountered was probably due to the exact=True pin.

jan-janssen · 2020-12-22T12:20:35Z

The reason why we need to pin boost explicitly is conda-forge/mshr-feedstock#18 and as you removed my changes we are basically back at #120 .

I know that adding boost-cpp is a work around but to me it seems like a reasonable solution and I would like to try if it fixes conda-forge/mshr-feedstock#23

jan-janssen · 2020-12-22T12:24:05Z

As the tests take an hour to run I opened another pull request with my previous suggestion #124

jan-janssen · 2020-12-22T12:26:59Z

@jan-janssen, I don't know why conda-smithy splits the build jobs for osx for each boost-cpp. But the linux build also creates two (boost-cpp-dependent) builds for fenics-libdolfin. Any downstream issue you encountered was probably due to the exact=True pin.

It seems to be an issue with combined recipes. When there is just the lib package it creates two builds for linux as well as it should. The easiest way I was able to force two builds was by adding boost-cpp to each sub package.

mbargull · 2020-12-22T12:41:19Z

The reason why we need to pin boost explicitly is conda-forge/mshr-feedstock#18 and as you removed my changes we are basically back at #120 .

I know that adding boost-cpp is a work around but to me it seems like a reasonable solution and I would like to try if it fixes conda-forge/mshr-feedstock#23

That workaround is no good, because the pinning is too specific and is likely to cause downstream (build and user-site) issues.
conda-forge/mshr-feedstock#18 is about a problem with a boost-cpp version change from 1.70 to 1.72. I don't see any indication that the exact boost-cpp build is of significance there.

@jan-janssen, I don't know why conda-smithy splits the build jobs for osx for each boost-cpp. But the linux build also creates two (boost-cpp-dependent) builds for fenics-libdolfin. Any downstream issue you encountered was probably due to the exact=True pin.

It seems to be an issue with combined recipes. When there is just the lib package it creates two builds for linux as well as it should. The easiest way I was able to force two builds was by adding boost-cpp to each sub package.

The behavior for Linux I understand. It should be fine to build multiple variants in one jobs.
conda-smithy's behavior macOS I don't understand.

mbargull · 2020-12-22T13:12:44Z

recipe/meta.yaml

+# NOTE: Top-level environment with boost-cpp is only to separate the build for
+#       multiple boost versions into different CI jobs.
+# TODO: Needs investigation why "separate the two boost builds more strictly"
+#       would be necessary. Ref:
+#       https://github.com/conda-forge/mshr-feedstock/pull/23#issuecomment-749520383
+requirements:
+  host:
+    - boost-cpp


@jan-janssen, since you weren't happy with my changes, I added this.
This is still just a workaround. I don't know if it's necessary. If it is, the cause should be investigated.
Prime andidates for causes are always:

reused CMake caches (but build-libdolfin.sh seems to work in a fresh build dir already, so maybe not this)

unexpected behavior from conda-build with outputs, e.g., PIP_NO_DEPENDENCIES & co not set for outputs/*/script conda/conda-build#3993

mbargull · 2020-12-22T13:12:59Z

@conda-forge-admin, please rerender

…da-forge-pinning 2020.12.22.11.23.16

jan-janssen · 2020-12-22T13:48:11Z

@mbargull I am fine with the changes - I was not aware of the requirements in addition to outputs. So if you agree I would suggest to merge this one once the tests passed.

mbargull · 2020-12-22T14:33:33Z

I was not aware of the requirements in addition to outputs.

Yeah, conda-build has a build stage before the builds of the outputs take place. Which means you can create a build/host env combo in which build.sh/bld.bat (if existing) will be run. Here we just create and don't use it -- meaning it's useless, but it forced boost-cpp to be in the top-level variant matrix to persuade conda-smithy to split the jobs.

So if you agree I would suggest to merge this one once the tests passed.

Sure, go ahead if it looks fine to you! Just trying to help with avoiding subsequent issues here ;).

mbargull · 2020-12-22T14:48:37Z

If we want to be super-correct, it looks like boost-cpp should also be a host+run dependency for fenics-dolfin: https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=255038&view=logs&j=9a864fd9-6c8f-52ca-79ce-2aa6dca1a1de&t=10fc5aa2-324e-5982-4c88-6b31fcab16b3&l=3300
, i.e., it's a direct dependency.
(In practice, due to the {{ pin_subpackage("fenics-libdolfin", exact=True) }}, the dep will be correctly installed.)

@jan-janssen, you seemed to be time-constrained here, so I'll leave this up to you if you want to fix this now or rather later.

jan-janssen · 2020-12-22T14:59:21Z

@jan-janssen, you seemed to be time-constrained here, so I'll leave this up to you if you want to fix this now or rather later.

I open another pull request for this and merge this one so we can continue with conda-forge/mshr-feedstock#23

Update meta.yaml

4ef26c6

jan-janssen requested review from garth-wells, johannesring, mikaem and minrk as code owners December 22, 2020 09:20

conda-forge-linter and others added 2 commits December 22, 2020 09:22

MNT: Re-rendered with conda-build 3.20.5, conda-smithy 3.8.5, and con…

24bcdb9

…da-forge-pinning 2020.12.21.13.42.08

Update meta.yaml

c3f4148

jan-janssen and others added 2 commits December 22, 2020 10:37

Update meta.yaml

3b52d2b

MNT: Re-rendered with conda-build 3.20.5, conda-smithy 3.8.5, and con…

020488b

…da-forge-pinning 2020.12.21.13.42.08

jan-janssen mentioned this pull request Dec 22, 2020

Combine multiple migrations conda-forge/mshr-feedstock#23

Merged

5 tasks

jan-janssen and others added 2 commits December 22, 2020 11:49

Increase build number

ebc15ac

Fix boost-cpp deps, use hdf5 dep from run_exports

0b47bdf

mbargull reviewed Dec 22, 2020

View reviewed changes

MNT: Re-rendered with conda-build 3.20.5, conda-smithy 3.8.5, and con…

aa050ff

…da-forge-pinning 2020.12.22.11.23.16

mbargull reviewed Dec 22, 2020

View reviewed changes

Separate CI jobs for boost-cpp versions

a844a73

mbargull reviewed Dec 22, 2020

View reviewed changes

MNT: Re-rendered with conda-build 3.20.5, conda-smithy 3.8.5, and con…

6cf0bfd

…da-forge-pinning 2020.12.22.11.23.16

jan-janssen mentioned this pull request Dec 22, 2020

Add boost-cpp to each package to trigger two builds for linux (1.72 and 1.74) #124

Closed

5 tasks

jan-janssen merged commit 697f470 into conda-forge:master Dec 22, 2020

jan-janssen mentioned this pull request Dec 22, 2020

Add boost-cpp as dependency for fenics-dolfin #125

Merged

5 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Try to trigger new boost build #123

Try to trigger new boost build #123

jan-janssen commented Dec 22, 2020

conda-forge-linter commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

github-actions bot commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

github-actions bot commented Dec 22, 2020

mbargull commented Dec 22, 2020

mbargull Dec 22, 2020

jan-janssen Dec 22, 2020

mbargull Dec 22, 2020

mbargull Dec 22, 2020

mbargull Dec 22, 2020

mbargull commented Dec 22, 2020 •

edited

Loading

jan-janssen commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

mbargull commented Dec 22, 2020

mbargull Dec 22, 2020

jan-janssen Dec 22, 2020

mbargull commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

mbargull commented Dec 22, 2020

mbargull commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

Try to trigger new boost build #123

Try to trigger new boost build #123

Conversation

jan-janssen commented Dec 22, 2020

conda-forge-linter commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

github-actions bot commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

github-actions bot commented Dec 22, 2020

mbargull commented Dec 22, 2020

mbargull Dec 22, 2020

Choose a reason for hiding this comment

jan-janssen Dec 22, 2020

Choose a reason for hiding this comment

mbargull Dec 22, 2020

Choose a reason for hiding this comment

mbargull Dec 22, 2020

Choose a reason for hiding this comment

mbargull Dec 22, 2020

Choose a reason for hiding this comment

mbargull commented Dec 22, 2020 • edited Loading

jan-janssen commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

mbargull commented Dec 22, 2020

mbargull Dec 22, 2020

Choose a reason for hiding this comment

jan-janssen Dec 22, 2020

Choose a reason for hiding this comment

mbargull commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

mbargull commented Dec 22, 2020

mbargull commented Dec 22, 2020

jan-janssen commented Dec 22, 2020

mbargull commented Dec 22, 2020 •

edited

Loading