Closed bergmark closed 8 years ago
eventstore
has been updated and uploaded on Hackage
Mandrill 0.5.1.0 is on Hackage.
A reminder that when there are lots of packages pending like here there's no need to comment on individual updated packages as it doesn't affect our workflow. No sweat :)
@bergmark I guessed that, sorry! It would be quite handy for us to be able to toggle somehow the checkboxes generated automatically, once the relevant packages we maintain are patched, but apparently we lack the permission to do so. That might cut comment traffic quite a bit! :wink:
I’ve never been told about that .:!
. I need to inquire a little bit deeper before pushing a support fix. :)
I should also note this list of packages that have known build failures:
@bergmark: Where are you seeing these build failures? Are these via matrix.hacakge.* or via the stackage builder?
@brendanhay we build the whole stackage with new aeson already (--allow-newer
kind of way)
@bergmark Hi, I'm the maintainer of argon. In my cabal file I have aeson >= 0.8
. What should I do exactly?
@rubik there's a compilation error you need to fix.
smoothie-0.4.2.2
just added support for aeson-0.11
.
Upped microlens-aeson
upper limit for aeson
.
Can you include instructions on how to build with the newer aeson on stack? stack build
doesn't work with resolver: nightly-2016-02-10
, it still picks aeson 0.9.0.1.
You can add an extra-dep to your stack.yaml file, e.g.:
extra-deps:
On Wed, Feb 10, 2016 at 10:05 AM, Johan Tibell notifications@github.com wrote:
Can you include instructions on how to build with the newer aeson on stack? stack build doesn't work with resolver: nightly-2016-02-10, it still picks aeson 0.9.0.1.
— Reply to this email directly or view it on GitHub https://github.com/fpco/stackage/issues/1211#issuecomment-182246744.
ekg and ekg-json done.
argon done. New version (0.4.1.0) pushed to Hackage.
elm-bridge is done
haskell-neo4j-client is done, new version 0.3.2.1 pushed to hackage
new version hoauth2-0.5.3
pushed to hackage.
The package twitter-feed-0.2.0.5 just uploaded to Hackage.
userid, happstack-authenticate, clckwrks, and clckwrks-plugin-page should all be updated and on hackage.
Thanks
A couple of new ones:
Can the new fresh list be generated? There is a lot of noise, and new ping might be a good idea.
riak is fixed http://hackage.haskell.org/package/riak-1.0.0.1
There are now two packages that have been disabled because they need aeson 0.11; bloodhound and katip-elasticsearch
I see pandoc.git allows aeson-0.11 now. Pandoc looks like the biggest blocker currently in terms of revdeps. Maybe @jgm can comment on plans for the next pandoc release?
Also servant-server and webdriver have some number of revdeps: and probably Michael's packages also I imagine but didn't check.
Probably good to run stackage-curator again to check for any new issues like:
which first appeared in #1261.
I released ratel-0.1.1 10 days ago with support for newer versions of Aeson.
Just released servant-0.4.4.7
which allows aeson-0.11.*
.
We're left with under a dozen packages on the naughty list. Please be advised that Stackage nightly builds will be switching over to aeson-0.11 soon, and any offending packages will be temporarily disabled from the build.
aeson-0.11.1.1 depended on by:
Made revisions for haxl
and servant-yaml
Thanks - down to 9 packages now. :)
I've just released pandoc 1.17, allowing aeson 0.11. Hope this allows you to move forward.
+++ Jens Petersen [Mar 17 16 03:45 ]:
I see pandoc.git allows aeson-0.11 now. Pandoc looks like the biggest blocker currently in terms of revdeps. Maybe [1]@jgm can comment on plans for the next pandoc release?
Just to not forget, latest pandoc-citeproc allows aeson 0.11 as well
ihaskell and ipython-kernel should be fixed
I just merged and released version 0.1.8 yesod-auth-oauth2 allowing aeson 0.11. Thanks, @k-bx.
Fixed dixi
Wow, thanks Guys! - this is really great. We're down to just a few now. :-)
aeson-0.11.1.1 is out of bounds for:
Additionally (#1296) pandoc-1.17 needs 3 other packages to be updated.
Also argon is also going to cause problems #1215, but we could disable it if needed.
datadog, flowdock, and hpc-coveralls also don't have any consumers in Stackage, so I think we could disable them if needed for moving to aeson-0.11.
Latest dixi requires servant-0.5 (#1290).
Current plan is to wait a little more for BlogLiterately-diagrams to support pandoc-1.17, and then move to aeson-0.11 and pandoc-1.17, and temporarily disable hpc-coveralls, dixi (sorry, because of needing servant-0.5), datadog, flowdock, and argon. Of course they can be re-enabled as soon as they support aeson-0.11. Hope this plan sounds okay.
Why does servant-0.5 lead to dixi being disabled? Is it simply not being included in stackage yet?
@liamoc we currently have a constraint on servant see #1290 - so we would re-enable dixi as soon as servant-0.5 is ready for Stackage Nightly. (Edit: Alternative would be to disable some of the servant consumers that have not been updated yet to 0.5, which may still happen depending on how long they take...)
OK, that's fine. It's not possible for dixi to work on both 0.4 and 0.5, so this solution seems best.
Okay thanks for understanding.
Leaving this open still since there could still be some build issues as Adam also mentioned earlier.
:relieved:
Thank you everyone for all the work on this!
Should be in 2016-03-24 Nightly, which is being pushed now.
https://github.com/guillaume-nargeot/hpc-coveralls is now fixed in version 1.0.4. Sorry for the delay.
Please see the changelog. Most likely your package will be fine without any changes but there are subtle semantic changes from 0.10.
Please let us (stackage curators and aeson maintainers) know if there are any issues with this upgrade as soon as possible. For the 0.10 release the maintainer was unresponsive, we didn't consistently poke stackage maintainers enough, and a lot of issues didn't show up until months later. I hope we can do better on all three fronts this time!
Here is some info on building nightly against aeson 0.11: https://github.com/bos/aeson/issues/355
aeson-0.11.0.0 is out of bounds for: