[GHC] #9176: GHC not generating dyn_hi files

[GHC]

#9176: GHC not generating dyn_hi files
-------------------------------------+-------------------------------------
        Reporter:  heatsink          |                Owner:
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  7.9
      Resolution:                    |             Keywords:  dynamic
Operating System:  Unknown/Multiple  |         Architecture:  x86_64
                                     |  (amd64)
 Type of failure:  Other             |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by ezyang):

 So, let's see if we can get this straight:

 1. In principle, it would be better if we didn't have any `dyn_hi` files.
 We should only have an `hi` file which everyone uses for typechecking.
 Because `-dynamic` does not affect type-checking/core optimization (it
 only affects the backend), any `dyn_hi` file you generate SHOULD be the
 same as `hi`. I'm pretty sure there isn't anything that is affected by
 this, and if there IS something we should kill it.

 2. In practice, if `-dynamic-too` does not work (as claimed on Windows),
 you have to run the compiler twice (once with `-dynamic` and once without)
 to actually get both dynamic and normal object files. Now, GHC is
 nondeterministic, so we're not guaranteed to get the same interface each
 time. That's trouble. Additionally, if you don't call GHC with exactly the
 same flags as the first time, you will end up with different interface
 files easily.

 The moral of the story is that you should not be allowed compile static
 objects, and THEN compile dynamic objects in a separate GHC invocation.
 Then the nondeterminism problem goes away and we can kill `dyn_hi` files
 with fire.

 It seems to me that `-dynamic-too`, even for platforms which "don't
 support it". I am not exactly sure how `-dynamic-too` is implemented, but
 in the worst case scenario it should be sufficient to just run the entire
 backend pipeline twice.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9176#comment:11>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler