#14081: 8.2.1 runghc from Windows x32 segfaults an all programs ----------------------------------+------------------------------ Reporter: sergv | Owner: (none) Type: bug | Status: new Priority: highest | Milestone: 8.2.2 Component: Compiler | Version: 8.2.1 Resolution: | Keywords: Operating System: Windows | Architecture: x86 Type of failure: Runtime crash | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | ----------------------------------+------------------------------ Comment (by sergv): I still wasn't able to use gdb to any good, but I think I found out the problem. The tl;dr version is: `addDLLHandle` assumes that import table is always present and it is not the case for 32-bit `ntdll.dll`. The long version is: I have stumbled upon some code for reading import table - https://stackoverflow.com/questions/15960437/how-to-read-import-directory- table-in-c#17457077. The GHC currently uses somewhat different but simpler approach. In particular it doesn't use anything resembling `Rva2Offset` function. I tried to use `Rva2Offset` and friends from the post but it didn't work. I didn't manage to debug it, but while trying to I noticed that post's code explicitly checks for the case when dll has no import table. I added this check to ghc and it seems that was enough to get `ghc --interactive` working. The check is: {{{ diff --git i/rts/linker/PEi386.c w/rts/linker/PEi386.c index 42e700805e..011b0a8314 100644 --- i/rts/linker/PEi386.c +++ w/rts/linker/PEi386.c @@ -240,6 +240,13 @@ static void addDLLHandle(pathchar* dll_name, HINSTANCE instance) { (PIMAGE_IMPORT_DESCRIPTOR)((BYTE *)instance + header-> OptionalHeader.DataDirectory[IMAGE_DIRECTORY_ENTRY_IMPORT].VirtualAddress); + bool importTableMissing = + header->OptionalHeader.DataDirectory[IMAGE_DIRECTORY_ENTRY_IMPORT].Size == 0; + + if (importTableMissing) { + return; + } + /* Ignore these compatibility shims. */ const pathchar* ms_dll = WSTR("api-ms-win-"); const int len = wcslen(ms_dll); }}} @Phyx- is the fix sensible? Should we try to merge it in? -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14081#comment:20 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler