[GHC] #8935: Obscure linker bug leads to crash in GHCi

[GHC]

#8935: Obscure linker bug leads to crash in GHCi
-------------------------------------+------------------------------------
        Reporter:  simonmar          |            Owner:  simonmar
            Type:  bug               |           Status:  new
        Priority:  high              |        Milestone:  7.8.3
       Component:  Runtime System    |          Version:  7.8.1-rc2
      Resolution:                    |         Keywords:
Operating System:  Unknown/Multiple  |     Architecture:  Unknown/Multiple
 Type of failure:  GHCi crash        |       Difficulty:  Rocket Science
       Test Case:                    |       Blocked By:
        Blocking:                    |  Related Tickets:
-------------------------------------+------------------------------------
Changes (by trommler):

 * status:  infoneeded => new


Comment:

 Replying to [comment:29 simonmar]:
 > My apologies for making a buggy test program :-)  Thanks for eliminating
 the strangeness there.
 No problem! I learned a lot about dynamic linking!
 > So it seems that `dlsym` always returns a bogus result for `environ`.
 No, no, it does not look bad at all. The address looks strange but that is
 caused by the fact that it was created by the dynamic linker. Look at
 `*defl_env` and `*env` in the second program in comment:25. The addresses
 match and I checked they really point to the environment array!

 In the first program the static linker creates the reference for environ
 because we take its address
 `&environ` to print it. So there is a static symbol already there and that
 confuses the dynamic linker when it searches in `libgmp` and then finds
 `environ` in `libc`.

 Once we have fixed this bug I will read the ELF specs again and see if
 giving us a defunct symbol when we load a weak symbol again is really
 allowed by the specification.
 >
 > Your proposed fix only works if we load libraries with `RTLD_LOCAL`
 rather than `RTLD_GLOBAL`, I think.  That is probably fine, but it needs
 testing.  This commit that introduced `internal_dlsym` explains the
 rationale: be497c202b790999c3fd0ddc4a4176b8cf6acf7e
 >
 > I'm fine with refactoring away the `hdl` argument.
 There is a branch on my github
 [https://github.com/trommler/ghc/tree/rtld-8935] where I implemented that
 and what I proposed above. While my AMD 64 bit and my old PowerMac G5
 build ghc, I will read up on the rationale for `internal_dlsym`.

 If validate succeeds and I am sufficiently confident I understand what
 `internal_dlsym` needs to do. I'll post my patches.

 We can remove the info needed.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8935#comment:30>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler