Closed phadej closed 7 years ago
![Uploading FireShot Capture - amazonka-core - packag_ - http___matrix.hackage.haskell.org_package_amazonka-core-before.png…]()
Via this package, virtually all amazonka-*
packages are non-buildable, e.g.
Reported on Stackage issue, DAV-1.2
:
[2 of 2] Compiling Network.Protocol.HTTP.DAV ( Network/Protocol/HTTP/DAV.hs, dist/build/Network/Protocol/HTTP/DAV.o )
Network/Protocol/HTTP/DAV.hs:160:32: error:
Not in scope: data constructor `StatusCodeException'
Perhaps you meant variable `matchStatusCodeException' (line 160)
Perhaps you want to add `StatusCodeException' to the import list
in the import of `Network.HTTP.Client'
(Network/Protocol/HTTP/DAV.hs:77:1-169).
judging from http://packdeps.haskellers.com/reverse/http-client, there's lots of packages depending on http-client directly whose authors forgot to put upper bounds in place, and sadly this compounds via indirect dependencies :-/
I'll add http-client
tag on matrix.h.h.o
for packages I queued to check this issue
before http-client-0.5 released:
after:
I'm correcting bounds atm.
Seems that http-client-0.5
itself fails to build with some configurations https://github.com/snoyberg/http-client/issues/210
kraken had two failures as well.
Not sure, seems (luckily!) that upper bounds used by many authors act as vaccination in masses: it helps prevent invalid plans even some packages aren't specifying them.
@phadej it's still fragile, because as soon as a 3rd party package relaxes its upper bound because it added support, suddenly install-plans will start to fail... the cabal system is designed to capture accurate information via direct dependencies (rather than relying on transitive dependencies to provide constraints), failure to do that breaks the system by making it fragile and hard to maintain.
More importantly, matrix.h.h.o is build upon the assumption that direct dependencies are expected to capture all necessary constraints. If this doesn't hold for 99% we may as well declare matrix.h.h.o death on arrival.
Some additional indirect fallout @ https://github.com/fpco/stackage-curator/issues/22 / https://github.com/fpco/stackage-install/issues/6
Stumbled over a few other ones:
http-reverse-proxy
keter
http-client-tls
For open ended issues like this, we can add comments for new entries but how about collecting everything in a checklist in the top comment (since we all can make edits to it)? That way we know what we need to do, and we can close when everything is done and re-open if new entries pop up.
@bergmark edited the first comment. Cannot verify as matrix is unresponsive to me.
I fixed the remaining packages
The stackage issue: https://github.com/fpco/stackage/issues/1676 Around 20 packages from Stackage was hold back due bounds.
Let's document packages which are or were broken due missing upper bounds.
I suspect that turn red too (waiting for matrix.h.h.o)
gogol-core
wreq
Also
http-conduit
breaks packages in the same way. let's check based http://packdeps.haskellers.com/reverse/http-conduit on:git-annex
:hoogle
http-proxy
Broken packages