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

Update pinnings #175

Merged
merged 1 commit into from
Oct 7, 2016
Merged

Update pinnings #175

merged 1 commit into from
Oct 7, 2016

Conversation

ocefpaf
Copy link
Member

@ocefpaf ocefpaf commented Jun 20, 2016

@pelson I am not really sure if these are needed.

@ocefpaf ocefpaf changed the title Add netcdf-cxx4 and netcdf-fortran Update pinnings Jun 28, 2016
@ocefpaf
Copy link
Member Author

ocefpaf commented Jun 28, 2016

I used this PR to add netcdf-cxx4 and netcdf-fortran, and update the pinnings to match what we have as latest. Merge with caution as some discussion is needed before we enforce this pinnings.

@jjhelmus
Copy link
Contributor

I'm in favor of these, although I am only worried about updates to the xz and sqlite pins.

A few notes:

  • readline 6.3 is out and available in the conda-forge channel but defaults is still pinning 6.2
  • tk 8.6.5 is out but the latest version in the conda-forge channel is 8.5.19. The defaults channel seems to be pinning 8.5.18 which is the latest version available from that channel and is in agreement with the current and proposed pinnings.
  • ncurses 6.0 is out but both the conda-forge and defaults channels provide 5.9.

@jakirkham
Copy link
Member

We were trying to get tk 8.6.5 done, but it is having issues ( conda-forge/tk-feedstock#1 ).

I think we should investigate the implications of ncurses 6.0 (degree of breakage). There are also some packages that are not pinned correctly floating around in conda-forge. So, I held off. ( conda-forge/ncurses-feedstock#14 ) The feedstocks should all be pinned correctly now or at least have been provided the opportunity.

@@ -37,14 +37,16 @@
'icu': 'icu 56.*',
'jpeg': 'jpeg 9*',
'libnetcdf': 'libnetcdf 4.4.*',
'libpng': 'libpng >=1.6.21,<1.7',
'libpng': 'libpng 1.6.23',
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Was there an API change which broke libpng in 23 as well? 😢

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not sure to be honest. But I got some reports of breaks in libpng 1.6.21 (conda-forge) and 1.6.22 (defaults)

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could you please provide references?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please see this issue: conda-forge/matplotlib-feedstock#41

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please see here too: conda-forge/libpng-feedstock#10

@pelson
Copy link
Member

pelson commented Jun 30, 2016

I'm confident in saying that this will tease out futher issues with the conda-forge stack (which other pinned version changes haven't! 😉). Though I'm in two minds as to whether I want to aim for some stability for the next few weeks (SciPy is on, and I'll be there -> lots of new users, and me not being fully available to fix any problems).

@pelson
Copy link
Member

pelson commented Jun 30, 2016

In principle, I'm very happy with moving these forwards. In practice, I could see an argument for doing this now, but can also see an argument for getting some stability. Are there drivers for doing this to solve existing issues (and do they outweigh the benefit of a few weeks of stability)?

@ocefpaf
Copy link
Member Author

ocefpaf commented Jun 30, 2016

SciPy is on, and I'll be there -> lots of new users, and me not being fully available to fix any problems

By no means I want this merge soon nor MTN PRs updating this. Maybe we should only do that once we figure out a better way to release new packages in an unstable channel first.

Are there drivers for doing this to solve existing issues?

Mostly is to update and move forward. However, xz is now a real problem as defaults moved to 5.2.* and xz is tied to Python 3.x. Meaning that the solver is updating Python and xz from defaults and breaking/not-resolving everything that is pinned to xz 5.0.*.

@pelson
Copy link
Member

pelson commented Jun 30, 2016

Thanks @ocefpaf. We can certainly separate updating xz. That does sound like an important step to keep up to speed with the latest anaconda.

@ocefpaf
Copy link
Member Author

ocefpaf commented Jun 30, 2016

Thanks @ocefpaf. We can certainly separate updating xz

See conda-forge/staged-recipes#907 (comment) for one example where that is important. (I can prepare another PR here just xz later today if you don't beat me to it.)

@jjhelmus
Copy link
Contributor

The defaults channel's python-3.x also pins a more recent sqlite than our current pinnings (3.13.* vs 3.9.*)

From the os-x python-3.5.1-5 package:

        - bzip2
        - zlib 1.2.*
        - sqlite 3.13.*
        - readline 6.2
        - tk 8.5.18
        - openssl 1.0.2*
        - xz 5.2.*

I would be happy with a PR that just updated xz and sqlite. A conda-forge Python 3.5 recipe is already read with these versions, conda-forge/python-feedstock#26

@ocefpaf
Copy link
Member Author

ocefpaf commented Jun 30, 2016

The defaults channel's python-3.x also pins a more recent sqlite than our current pinnings (3.13.* vs 3.9.*)

Agreed. See #182

@jakirkham
Copy link
Member

  • readline 6.3 is out and available in the conda-forge channel but defaults is still pinning 6.2

Should have added we also pin to 6.2.

@jjhelmus
Copy link
Contributor

jjhelmus commented Jul 19, 2016

Should have added we also pin to 6.2.

defaults seems to be pinning to 6.2 also so this is not an issue. Although trying to upgrade readline in a conda environment can have some interesting results (conda upgrade --all works fine):

$ conda upgrade -c conda-forge readline
Using Anaconda Cloud api site https://api.anaconda.org
Fetching package metadata .........
Solving package specifications: ..........

Package plan for installation in environment /home/jhelmus/anaconda/envs/readline_test:

The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    python-1.2                 |                0         773 KB  conda-forge
    readline-6.3               |                1         787 KB  conda-forge
    ------------------------------------------------------------
                                           Total:         1.5 MB

The following packages will be UPDATED:

    readline: 6.2-0   conda-forge --> 6.3-1 conda-forge

The following packages will be DOWNGRADED due to dependency conflicts:

    python:   3.5.2-1 conda-forge --> 1.2-0 conda-forge

Proceed ([y]/n)? 

@jakirkham
Copy link
Member

The following packages will be DOWNGRADED due to dependency conflicts:

   python:   3.5.2-1 conda-forge --> 1.2-0 conda-forge

😱

@jakirkham
Copy link
Member

Has some merge conflicts that need to be resolved.

@pelson
Copy link
Member

pelson commented Sep 30, 2016

@ocefpaf - would you mind rebasing this - I think we can go ahead and merge.

@ocefpaf
Copy link
Member Author

ocefpaf commented Sep 30, 2016

I am on the road but I'll try to connect later today and rebase. (It is
almost a miracle to get 3G coverage from where I am today :-)

@@ -39,10 +39,12 @@
'libblitz': 'libblitz 0.10|0.10.*',
'libmatio': 'libmatio 1.5.*',
'libnetcdf': 'libnetcdf 4.4.*',
'libpng': 'libpng >=1.6.21,<1.7',
'libpng': 'libpng >=1.6.23,<1.7',
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is more restrictive (an working in some cases) than the current setup. So I am merging this to get the other changes in too. We have another PR that pins libpng to the bone and any further discussion should go there.

See #211

@ocefpaf
Copy link
Member Author

ocefpaf commented Sep 30, 2016

@ocefpaf - would you mind rebasing this - I think we can go ahead and merge.

Done. I will be online for a couple of hours. Let me know if you need anything else.

@183amir
Copy link
Contributor

183amir commented Sep 30, 2016

Hey guys, please please pin libpng on a patch number. I beg of you. I have been having trouble with this for so long but you don't listen. Please pin it on 1.6.22 too to be inline with the defaults channel.

@ocefpaf
Copy link
Member Author

ocefpaf commented Sep 30, 2016

Hey guys, please please pin libpng on a patch number. I beg of you. I have been having trouble with this for so long but you don't listen. Please pin it on 1.6.22 too to be inline with the defaults channel.

There is no need for desperation. See #211 (comment)

The context of this PR here is not to fix that. I am just waiting for @pelson's comment on #211 (comment) to rebase and pin there.

@183amir
Copy link
Contributor

183amir commented Sep 30, 2016

Here is a log that shows matplotlib compiled with 1.6.24 does not work with libpng 1.6.23 https://travis-ci.org/conda-forge/bob.ip.facedetect-feedstock/jobs/163096191

@ocefpaf ocefpaf merged commit f7fbf72 into conda-forge:master Oct 7, 2016
@ocefpaf ocefpaf deleted the update_pins branch October 7, 2016 19:16
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Development

Successfully merging this pull request may close these issues.

5 participants