[GHC] #14558: Unable to parse integer-gmp's Cabal file

[GHC]

#14558: Unable to parse integer-gmp's Cabal file
-------------------------------------+-------------------------------------
        Reporter:  taylorfausak      |                Owner:  hvr
            Type:  task              |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Core Libraries    |              Version:  8.2.2
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by taylorfausak):

 Come on y'all. I'm trying to avoid turning this into an us versus them
 argument for all values of us and them. The problem described in this
 ticket affects people trying to install GHC >= 8.2 with Stack < 1.6.1. It
 also affects people like me who want to parse the package descriptions of
 the core libraries without depending on bleeding edge Cabal features.

 I'm also trying to avoid arguing about the `^>=` operator itself. We have
 a new operator, it's been talked to death, people are going to use it.
 Currently `^>= x.y.z` is syntactic sugar for `>= x.y.z && < x.(y+1)`. In
 the future, Hackage trustees may be able to make changes to `^>=`
 constraints because they are "soft".

 That brings me once again to the crux of this ticket:

 > it's frustrating that a core library is using a bleeding edge feature
 for basically no reason

 Is there a compelling reason for `integer-gmp` to use `^>=`? It has
 already been demonstrated that using `^>=` causes problems. I will happily
 close the ticket and be on my merry way if someone can say: "Yes,
 `integer-gmp` (or any core library) needs `^>=` (or any Cabal 2 feature)
 because //X//." I have not heard such a reason yet. It seems to me that
 the GHC developers are refusing to make a simple fix that will solve a
 problem for a large number of users.

 Also I'm having a hard time wrapping my head around why the `^>=` operator
 would ever be necessary for core libraries. As many people have pointed
 out in this ticket, `integer-gmp` cannot be reinstalled. So any revisions
 to its package description are essentially ignored. The whole point of
 `^>=`, as far as I can tell, is to allow Hackage trustees to revise the
 bounds if new valid build plans are discovered. That means using `^>=` in
 `integer-gmp` only saves a few keystrokes at best.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14558#comment:30>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler