#11748: GHC runtime linker: fatal error: I found a duplicate definition for symbol -------------------------------------+------------------------------------- Reporter: jeiea | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 (Linking) | Resolution: | Keywords: Operating System: Windows | Architecture: x86_64 Type of failure: Compile-time | (amd64) crash | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Changes (by jeiea): * Attachment "ghc-bug-maybe.7z" added. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11748 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11748: GHC runtime linker: fatal error: I found a duplicate definition for symbol -------------------------------------+------------------------------------- Reporter: jeiea | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 (Linking) | Resolution: | Keywords: Operating System: Windows | Architecture: x86_64 Type of failure: Compile-time | (amd64) crash | Test Case: Blocked By: | Blocking: Related Tickets: #11223 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Changes (by Phyx-): * related: => #11223 Comment: The dependency list for the example is quite large, including lots of packages that haven't been updated for 8.0 yet. As such it's quite difficult to reproduce and check on HEAD. For now, I'll assume that this is a duplicate of #11223 since that should fix a lot of these duplicate symbols errors. Can you perhaps trim the amount of packages needed in the example? -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11748#comment:2 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11748: GHC runtime linker: fatal error: I found a duplicate definition for symbol -------------------------------------+------------------------------------- Reporter: jeiea | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 (Linking) | Resolution: | Keywords: Operating System: Windows | Architecture: x86_64 Type of failure: Compile-time | (amd64) crash | Test Case: Blocked By: | Blocking: Related Tickets: #11223 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Changes (by jeiea): * Attachment "ghc-bug-maybe-2.7z" added. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11748 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11748: GHC runtime linker: fatal error: I found a duplicate definition for symbol -------------------------------------+------------------------------------- Reporter: jeiea | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 (Linking) | Resolution: | Keywords: Operating System: Windows | Architecture: x86_64 Type of failure: Compile-time | (amd64) crash | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by jeiea): Followed your instruction after {{{stack clean}}}, still same error message occurred. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11748#comment:5 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11748: GHC runtime linker: fatal error: I found a duplicate definition for symbol -------------------------------------+------------------------------------- Reporter: jeiea | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 (Linking) | Resolution: | Keywords: Operating System: Windows | Architecture: x86_64 Type of failure: Compile-time | (amd64) crash | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Changes (by jeiea): * Attachment "result.txt" added. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11748 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11748: GHC runtime linker: fatal error: I found a duplicate definition for symbol -------------------------------------+------------------------------------- Reporter: jeiea | Owner: Type: bug | Status: closed Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 (Linking) | Resolution: duplicate | Keywords: Operating System: Windows | Architecture: x86_64 Type of failure: Compile-time | (amd64) crash | Test Case: Blocked By: | Blocking: Related Tickets: #11223 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Changes (by Phyx-): * status: new => closed * resolution: => duplicate * related: => #11223 Comment: Right, looking at it more closely, I think I understand the problem. The `Win32` package already provides a function named `rgb`. Due to the call `C.include ""` in `Lib.hs`. The TH seems to be producing another function called `rgb` and hence the duplicate symbols error. GHC 7.x uses the old runtime linker which greedily loads symbols. So even though you don't use any symbols from the `Win32` package. Because you're linking against it an error is given. The second issue is that it's linking against the full `Win32.o` object file that cabal produces instead of the `Win32.a` file. In any case, you can't solve this in 7.10.3, you need the fix of #11223. So this is a combination of things, but I'll mark it as a duplicate since for it to work you'll need GHC 8.0. Feel free to re-open it if that doesn't solve it. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11748#comment:8 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11748: GHC runtime linker: fatal error: I found a duplicate definition for symbol -------------------------------------+------------------------------------- Reporter: jeiea | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 (Linking) | Resolution: | Keywords: Operating System: Windows | Architecture: x86_64 Type of failure: Compile-time | (amd64) crash | Test Case: Blocked By: | Blocking: Related Tickets: #11223 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Changes (by jeiea): * status: closed => new * resolution: duplicate => -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11748#comment:10 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11748: GHC runtime linker: fatal error: I found a duplicate definition for symbol -------------------------------------+------------------------------------- Reporter: jeiea | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Template Haskell | Version: 7.10.3 Resolution: | Keywords: Operating System: Windows | Architecture: Type of failure: Compile-time | Unknown/Multiple crash | Test Case: Blocked By: | Blocking: Related Tickets: #11223 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Changes (by Phyx-): * component: Compiler (Linking) => Template Haskell * architecture: x86_64 (amd64) => Unknown/Multiple Comment: The linker didn't change between `8.0.1` and `8.0.2`. I think this is a general limitation of TH. At least the current implementation. The linker is correct in saying you have a duplicate symbol. One is coming from the code that `inline-c` generates, which ends up producing a symbol called `rgb` (probably from the macro `RGB` in `windows.h`). The second comes from the `Win32` package. The `8.0.2` linker should have explicitly pointed these out. The issue is that TH uses a version of `Win32` that has all the symbols in the same object file. So the link has no choice but to load it all, even if you don't use `rgb` because it's in the same object file. If instead TH would use the archive instead of the `whole-package-as-object-file` variant then it should work. I'll leave this open so someone who knows more about TH can comment on the issue. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11748#comment:12 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11748: GHC runtime linker: fatal error: I found a duplicate definition for symbol -------------------------------------+------------------------------------- Reporter: jeiea | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Template Haskell | Version: 7.10.3 Resolution: | Keywords: Operating System: Windows | Architecture: Type of failure: Compile-time | Unknown/Multiple crash | Test Case: Blocked By: | Blocking: Related Tickets: #11223 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by rwbarton): The original issue turns out to be simply due to the fact that `inline-c` depends on `Win32` (via `directory`), and your program's cabal file states a dependency on your local version `Win32a`. `Win32` defines functions like `rgb` as wrappers around Windows API macros and your clone `Win32a` also defines functions with the same names. So you do have duplicate definitions for symbols. The issue doesn't have anything to do with `inline-c` besides the fact that it is causing the real `Win32` to be linked in. The issue is that, unlike symbols that come from Haskell functions, we don't mangle the symbols defined in C source files with the package identifier. I think there is a ticket about changing this, but for now it's a known limitation that you have to ensure that symbol names in C files you link in globally unique. Using the symbol name `rgb` in `Win32` is really not very good, as it is fairly likely to collide with someone else's symbol, and we should change that. In the `unix` package, we use symbol names like `__hsunix_ptsname`. However using a more GHC-specific symbol name would not help here, since you copied the `Win32` source into your own package. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11748#comment:14 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11748: GHC runtime linker: fatal error: I found a duplicate definition for symbol -------------------------------------+------------------------------------- Reporter: jeiea | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Template Haskell | Version: 7.10.3 Resolution: | Keywords: Operating System: Windows | Architecture: Type of failure: Compile-time | Unknown/Multiple crash | Test Case: Blocked By: | Blocking: Related Tickets: #11223 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by rwbarton): In the second reproducer, `child1`'s module `Lib` calls `rgb`, which is supposed to be defined by `child1`'s `HsGDI.c`. But the library `child1` does not have `c-sources: src/HsGDI.c`, only the (irrelevant) `child1-exe` does. It happens that `ghc-bug-maybe-exe` links indirectly against `Win32` (via `inline-c`), which coincidentally would provide a symbol `rgb`, but because there is no dependency of `child1` on `Win32`, it happened that `Win32` was given before `child1` on the linker command line and so the linker didn't know that there would be a reference to `rgb` when it processed `Win32`. The right thing to do of course is to specify `c-sources: src/HsGDI.c` in the `child1` ''library'' section; then you might get a duplicate symbol error again, because you chose the same name as `Win32` did. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11748#comment:16 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11748: GHC runtime linker: fatal error: I found a duplicate definition for symbol -------------------------------------+------------------------------------- Reporter: jeiea | Owner: Type: bug | Status: closed Priority: normal | Milestone: Component: Template Haskell | Version: 7.10.3 Resolution: invalid | Keywords: Operating System: Windows | Architecture: Type of failure: Compile-time | Unknown/Multiple crash | Test Case: Blocked By: | Blocking: Related Tickets: #11223 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by rwbarton): By the way, if ghc-mod doesn't build with ghc-8.0.2, please open a new ticket! Linkers are remarkably complicated beasts and every linker error is a unique snowflake. Having the two issues in the original report and in comment:3 on this ticket was already a little bit confusing. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11748#comment:18 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11748: GHC runtime linker: fatal error: I found a duplicate definition for symbol -------------------------------------+------------------------------------- Reporter: jeiea | Owner: Type: bug | Status: closed Priority: normal | Milestone: Component: Template Haskell | Version: 7.10.3 Resolution: invalid | Keywords: Operating System: Windows | Architecture: Type of failure: Compile-time | Unknown/Multiple crash | Test Case: Blocked By: | Blocking: Related Tickets: #11223 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by RyanGlScott): BTW rwbarton, is #9351 the globally unique symbol names ticket you alluded to in comment:14? -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11748#comment:20 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler