#7718: ios patch no 8: adjustor pools --------------------------------+------------------------------------------- Reporter: StephenBlackheath | Owner: Type: feature request | Status: patch Priority: normal | Milestone: 7.8.1 Component: Compiler | Version: 7.7 Resolution: | Keywords: Os: Other | Architecture: arm Failure: None/Unknown | Difficulty: Unknown Testcase: | Blockedby: Blocking: 7724 | Related: --------------------------------+------------------------------------------- Comment(by StephenBlackheath): Reminding myself of the reason... I believe I was slightly wrong with what I just said: the 'exec' and 'write' pointer being the same is not the issue. Here's a more accurate explanation: On iOS, ffi_prep_closure_loc must be used to construct the closure (because that is the only way to get around Apple restrictions). It requires the exec and writable pointers to be exactly the ones returned by ffi_closure_alloc. No such limitation exists on Linux because we are creating our own closures. createAdjustor returns only the exec pointer, so this is all that freeExec can receive initially. I can use the same trick as on Linux to go from writable -> exec. However, I can't use the same trick for exec -> writable, because it requires me to write into the memory, and the exec pointer is not writable! Therefore the hash table is the only way. Note that the Linux code writes into the writable pointer and reads from the exec pointer. Therefore they MUST be the same address, or it would not retrieve the same value. The whole business of writing into the pointer and advancing it by one may be totally unnecessary on Linux. -- Ticket URL: http://hackage.haskell.org/trac/ghc/ticket/7718#comment:17 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler