Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

bump to version 0.4.3 #12

Merged
merged 3 commits into from
Oct 18, 2016
Merged

bump to version 0.4.3 #12

merged 3 commits into from
Oct 18, 2016

Conversation

jakevdp
Copy link
Contributor

@jakevdp jakevdp commented Oct 13, 2016

No description provided.

@conda-forge-linter
Copy link

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.

@jakirkham
Copy link
Member

Please re-render with conda-smithy version 1.3.2 before merging this or any other PR. Can do in this PR or in a separate one that is merged first.

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 13, 2016

I'm getting an error with conda-smithy 1.3.2:

jakesmac:vega-feedstock jakevdp$ conda smithy rerender
No circle token.  Create a token at https://circleci.com/account/api and
put it in ~/.conda-smithy/circle.token
No appveyor token. Create a token at https://ci.appveyor.com/api-token and
Put one in ~/.conda-smithy/appveyor.token
No anaconda token. Create a token via
  anaconda auth --create --name conda-smithy --scopes "repos conda api"
and put it in ~/.conda-smithy/anaconda.token
Fetching package metadata .......
Traceback (most recent call last):
  File "/Users/jakevdp/anaconda/bin/conda-smithy", line 6, in <module>
    sys.exit(conda_smithy.cli.main())
  File "/Users/jakevdp/anaconda/lib/python3.5/site-packages/conda_smithy/cli.py", line 239, in main
    args.subcommand_func(args)
  File "/Users/jakevdp/anaconda/lib/python3.5/site-packages/conda_smithy/cli.py", line 185, in __call__
    configure_feedstock.main(args.feedstock_directory)
  File "/Users/jakevdp/anaconda/lib/python3.5/site-packages/conda_smithy/configure_feedstock.py", line 444, in main
    render_run_docker_build(env, config, forge_dir)
  File "/Users/jakevdp/anaconda/lib/python3.5/site-packages/conda_smithy/configure_feedstock.py", line 44, in render_run_docker_build
    if not ResolvedDistribution(meta, pkgs).skip():
  File "/Users/jakevdp/anaconda/lib/python3.5/site-packages/conda_build_all/resolved_distribution.py", line 77, in __getattr__
    with setup_vn_mtx_case(self.special_versions):
  File "/Users/jakevdp/anaconda/lib/python3.5/contextlib.py", line 59, in __enter__
    return next(self.gen)
  File "/Users/jakevdp/anaconda/lib/python3.5/site-packages/conda_build_all/resolved_distribution.py", line 21, in setup_vn_mtx_case
    orig_npy = conda_build.config.config.CONDA_NPY
AttributeError: module 'conda_build.config' has no attribute 'config'

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 13, 2016

OK, looks like I had an old version of conda-build-all (0.11). Perhaps that requirement should be added to the conda-smithy recipe?

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 13, 2016

Rerender in #13. I'll rebase this PR once that is merged.

@jakirkham
Copy link
Member

Sorry I missed that. Glad you got it sorted though.

The situation is more that to use conda-build version 2.0.x one needs conda-build-all version 1.0.0 and the conda-smithy version 1.2.0 or greater. Basically, this error wouldn't have occurred with conda-build version 1.21.x and the old conda-build-all. Not sure the right way to specify that ATM as I don't think we have conditional versioning of this sort.

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 13, 2016

Appveyor is failing because when activating the environment, The input line is too long.

Seems related to this issue, and the fix there is basically "make your PATH variable shorter".

It seems fixing this would require modifying the AppVeyor scripts somehow... any ideas on that?

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 14, 2016

@pelson @ocefpaf would love to hear any solutions for this Appveyor failure...

Also, Travis has been Queued for nearly 24 hours now.

@ocefpaf
Copy link
Member

ocefpaf commented Oct 14, 2016

@pelson @ocefpaf would love to hear any solutions for this Appveyor failure...

Sorry @jakevdp but I have no idea on both cases. Travis-CI had its outages before, but never like this.

Maybe we can work around the AppVeyor error... @msarahan do you have any idea why this is affecting only PY27? I lost track of things but it seems like the feedstoscks are still downloading miniconda. I thought we changed that a long time ago to use the AppVeyor's pre-installed miniconda. Maybe that will help here. If I get some extra time I will take a look this weekend.


Edit: xref conda-forge/conda-forge.github.io#252

@jakirkham
Copy link
Member

IIUC there was some bad interaction between conda and AppVeyor. @msarahan made a new deployment of conda to fix this last night, but I guess it took some time to propagate through. Most builds that have been restarted since appear to be working.

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 14, 2016

Appveyor still failing on 2.7, due to the PATH being too long.

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 14, 2016

Looks like some difference in how paths are specified between http://repo.continuum.io/miniconda/Miniconda-latest-Windows-x86.exe and http://repo.continuum.io/miniconda/Miniconda3-latest-Windows-x86.exe

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 14, 2016

FYI, AppVeyor has miniconda built-in, so maybe conda-forge should be using their version? appveyor/ci#359

@ocefpaf
Copy link
Member

ocefpaf commented Oct 14, 2016

FYI, AppVeyor has miniconda built-in, so maybe conda-forge should be using their version? appveyor/ci#359

Like I said above

thought we changed that a long time ago to use the AppVeyor's pre-installed miniconda.

I do use it in my non-conda-forge builds and I do remember @jjhelmus implementing that here. But since conda-build-setup was introduced I lost track as things got buried into deep layers and harder to debug 😒

@jakirkham
Copy link
Member

FYI, AppVeyor has miniconda built-in, so maybe conda-forge should be using their version? appveyor/ci#359

We changed to AppVeyor's included Miniconda in staged-recipes. It was kind of a mess as it is pretty out-of-date in some cases. It was suggested that we do the same in feedstocks. However, the person suggesting it changed their mind due to some issues they encountered. So we skipped it.

Given using Miniconda is already a problem as we spend a couple of minutes just updating from defaults packages to conda-forge packages and sometimes (more frequently of late) run into connectivity issues by slamming Anaconda.org in this way, the general feeling is we should have our own installer to cutdown build time and limit this impact. The situation with using an outdated copy of Miniconda will undoubtedly be no different in this respect.

We packaged everything needed to do this. Someone even got the basics for an installer together during a Hackathon at SciPy. It was transferred to a repo here. Though there are a number of important issues we need to address before we can proceed. Help on any of these would be appreciated.

@ocefpaf
Copy link
Member

ocefpaf commented Oct 14, 2016

It is a pity you think that is messy, or worse, have our own installer. That kind "forkish" behavior is the one that drivers people away.

BTW the default AppVeyor miniconda works wonders in all my other projects. Including building with conda-build 2. All I do is conda update --all without the conda-forge channel of course since conda/conda-build from there is unreliable, and I get all latest versions of conda and conda-build ready to go.

Ping @asmeurer who is also puzzled with some the the weird decision in conda-forge in conda-forge/conda-forge-build-setup-feedstock#38

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 14, 2016

Is there any significant difference in what you get if you (a) download and install miniconda-latest or (b) use an old miniconda and run conda update conda?

@ocefpaf
Copy link
Member

ocefpaf commented Oct 14, 2016

It was kind of a mess

BTW this is what I use and I do not think it is messy:

environment:
  matrix:
    - PYTHON: "C:\\Miniconda-x64"
      PYTHON_VERSION: "2.7"
      PYTHON_ARCH: "64"
    - PYTHON: "C:\\Miniconda"
      PYTHON_VERSION: "2.7"
      PYTHON_ARCH: "32"

init:
  - "ECHO %PYTHON_VERSION% %MINICONDA%"

install:
  - "set PATH=%PYTHON%;%PYTHON%\\Scripts;%PATH%"

as it is pretty out-of-date in some cases.

That part is easily fixed by a conda update --all.

@ocefpaf
Copy link
Member

ocefpaf commented Oct 14, 2016

Is there any significant difference in what you get if you (a) download and install miniconda-latest or (b) use an old miniconda and run conda update conda?

Not to my knowledge. My Other auto-build conda packages projects work just fine that way.

@asmeurer
Copy link
Member

I had heard that Miniconda was going to change to install conda into a separate environment, but I just tested the latest Miniconda OS X installer and it didn't do that, so I'm not so sure.

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 15, 2016

Can someone restart the AppVeyor build here?

@ocefpaf
Copy link
Member

ocefpaf commented Oct 15, 2016

Just did. Btw you should have AppVeyor rights here to restart it too. Let me know if that is not working.

@jakirkham
Copy link
Member

The failure will be resolved by re-rendering with conda-smithy version 1.3.3.

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 17, 2016

Appveyor builds completed after 10 hours in the queue, but one hit a network error. Since there's no way for Appveyor to restart a single job I restarted them all. Sigh...

@jakirkham
Copy link
Member

but one hit a network error.

You're not alone. Hence the long queue. 😒

Trying to track this down with upstream. ( https://github.com/Anaconda-Platform/support/issues/78 )

Since there's no way for Appveyor to restart a single job I restarted them all.

Asked upstream for this feature. ( appveyor/ci#1114 )

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 18, 2016

Another single network error in Appveyor after half a day in the queue. I'll restart all the builds again, since that's the only recourse. This is seriously trying my patience...

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 18, 2016

Given the constraint these two trials put on the probability of failure and duration of the queue, I could model it as a multinomial process compute an expectation value for how long this will take to finish... but that would probably just depress me.

Or maybe I should just merge?

@ocefpaf
Copy link
Member

ocefpaf commented Oct 18, 2016

@jakevdp just merge it as you can restart it on master to get the missing package to be uploaded. You already know that the build will pass and that error is only network issue.

Unfortunately an OS X image update on Travis-CI scared some people and they issued hundreds of feedstocks updates that are not skipping CIs and, in most cases, are not even affected by the OS X iamge update 😒

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 18, 2016

OK!

@jakevdp jakevdp merged commit ab86906 into conda-forge:master Oct 18, 2016
@jakirkham
Copy link
Member

FWIW it seems the merge passed on the first try.

@jakevdp
Copy link
Contributor Author

jakevdp commented Oct 18, 2016

FWIW it seems the merge passed on the first try.

yep, I noticed that 😄

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants