#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 Phyx-):
But the big surprise was that this upload of integer-gmp to hackage broke stack's promise of reproducible builds. (Even using snapshots dated well before the upload of integer-gmp)
Is this not a problem of stack? Should it not have been the one to guarantee that if it can't parse a package it does not enter it's system? If it makes such a guarantee it's up to it validate changes to preserve such guarantee. I don't understand how a new package can break LTS releases. It's unrealistic to expect GHC to never change. We can guarantee backwards compatibility but not forward compatibility. What I am missing here is why this is such a big deal. Yes the change was made by GHC upstream, yes it broke stack nightlies. But is it not the the purpose of nightlies to test the bleeding edge anyway? This isn't the last time the format will change. Let's fix the tools to actually adhere to what is in the `cabal-version` fields and move on. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14558#comment:21 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler