[GHC] #9762: #8696 + #9012

[GHC]

#9762: #8696 + #9012
-------------------------------------+-------------------------------------
              Reporter:  rwbarton    |            Owner:
                  Type:  bug         |           Status:  new
              Priority:  normal      |        Milestone:
             Component:  Compiler    |          Version:  7.8.3
            Resolution:              |         Keywords:
      Operating System:              |     Architecture:  Unknown/Multiple
  Unknown/Multiple                   |       Difficulty:  Unknown
       Type of failure:              |       Blocked By:
  None/Unknown                       |  Related Tickets:  #8696, #9012
             Test Case:              |
              Blocking:              |
Differential Revisions:              |
-------------------------------------+-------------------------------------

Comment (by nh2):

 Some discussion from #ghc for readers who'd like context:


 '''nh2''': thoughtpolice, '''rwbarton''': I am getting the "recompile with
 -fPIC" message in my `ghc -M` based setup, which you discussed in
 https://ghc.haskell.org/trac/ghc/ticket/8696. If I get this message, does
 it really mean that I need -fPIC or that I'm calling something wrong?
 '''rwbarton''': you probably don't want -fPIC specifically, but rather
 whichever one of -shared and -dynamic is the one you want
 '''rwbarton''': however, we can't know without more context
 '''rwbarton''': or possibly something is just broken :)
 '''nh2''': so, in the end I don't want any dynamic linking at all, but I
 understand it that I need some form of it for GHCI to load modules during
 TH evaluation, right?
 carter: yeah
 '''rwbarton''': yes
 carter: at least for onw
 carter: i think luite  got his TH patch ready?
 '''nh2''': so I'm not sure which one of -shared and -dynamic I want for
 ghci to "just work"
 '''rwbarton''': you will need to build both ways
 '''nh2''': so I'm already compiling with -dynamic-too; in this case, what
 do I have to add?
 '''rwbarton''': well then we need to see a real error
 '''nh2''': http://lpaste.net/113689
 '''rwbarton''': oh, and you are using hpc too
 '''rwbarton''': I thought I fixed this...
 '''rwbarton''': what version of ghc?
 '''rwbarton''': that should be fixed in 7.8.3 I think
 '''nh2''': 7.8.3, unfortunately
 '''rwbarton''': hm
 '''rwbarton''': oh, right
 '''rwbarton''': I knew about this when I fixed it :)
 '''rwbarton''': but I wrote the test in the same form as the surrounding
 code
 '''rwbarton''': hpc and TH won't interact well currently
 '''rwbarton''': I think you are one of 3 people to use HPC
 '''nh2''': we are a 4 person team
 '''nh2''': define "won't interact well" between [works if you disable it
 somehow for the libs loaded in ghci .... and .... you got no chance] :)
 '''rwbarton''': basically your bug is the bastard child of #9012 and #8696
 '''rwbarton''': well if you could build the dynamic objects without HPC
 then it should work
 '''nh2''': do you reckon there is also a full solution to this? Also, is
 this now a bug, should I file one?
 '''rwbarton''': there's a bunch of trivial fixes to every individual
 problem of this form
 '''rwbarton''': each one also costs a little performance in the associated
 situation
 '''rwbarton''': I guess if you are building with hpc then you don't care
 about this level of overhead
 '''nh2''': is there a way to tell the -dynamic-too invocation that I don't
 want HPC only for the dynamic files? Or will I have to stop using
 -dynamic-too and create the dynamic ones manually to achieve that?
 '''rwbarton''': but really we should find a better way to deal with the
 underlying issue, since it does cost a measurable amount of performance on
 anything built dynamically
 '''rwbarton''': probably the latter

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9762#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler