Validate fails on armhf-linux

After apply the 'configure : LLVM and LD detection improvements' patch from #10234 (closed) validation of the ghc-7.10 branch fails with:

inplace/bin/dll-split compiler/stage2/build/.depend-v-dyn.haskell \
  "DynFlags" "Annotations ApiAnnotation Avail Bag BasicTypes Binary \
  BooleanFormula BreakArray BufWrite Class CmdLineParser CmmType CoAxiom \
  ConLike Coercion Config Constants CoreArity CoreFVs CoreSubst CoreSyn \
  CoreTidy CoreUnfold CoreUtils CostCentre DataCon Demand Digraph \
  DriverPhases DynFlags Encoding ErrUtils Exception ExtsCompat46 \
  FamInstEnv FastFunctions FastMutInt FastString FastTypes Fingerprint \
  FiniteMap ForeignCall Hooks HsBinds HsDecls HsDoc HsExpr HsImpExp \
  HsLit PlaceHolder HsPat HsSyn HsTypes HsUtils HscTypes IOEnv Id \
  IdInfo IfaceSyn IfaceType InstEnv Kind Lexeme ListSetOps Literal \
  Maybes MkCore MkId Module MonadUtils Name NameEnv NameSet OccName \
  OccurAnal OptCoercion OrdList Outputable PackageConfig Packages Pair \
  Panic PatSyn PipelineMonad Platform PlatformConstants PprCore PrelNames \
  PrelRules Pretty PrimOp RdrName Rules Serialized SrcLoc StaticFlags \
  StringBuffer TcEvidence TcRnTypes TcType TrieMap TyCon Type TypeRep \
  TysPrim TysWiredIn Unify UniqFM UniqSet UniqSupply Unique Util Var \
  VarEnv VarSet Bitmap BlockId ByteCodeAsm ByteCodeInstr ByteCodeItbls \
  CLabel Cmm CmmCallConv CmmExpr CmmInfo CmmMachOp CmmNode CmmUtils \
  CodeGen.Platform CodeGen.Platform.ARM CodeGen.Platform.ARM64 \
  CodeGen.Platform.NoRegs CodeGen.Platform.PPC CodeGen.Platform.PPC_Darwin \
  CodeGen.Platform.SPARC CodeGen.Platform.X86 CodeGen.Platform.X86_64 \
  FastBool Hoopl Hoopl.Dataflow InteractiveEvalTypes MkGraph PprCmm \
  PprCmmDecl PprCmmExpr Reg RegClass SMRep StgCmmArgRep StgCmmClosure \
  StgCmmEnv StgCmmLayout StgCmmMonad StgCmmProf StgCmmTicky StgCmmUtils \
  StgSyn Stream"
compiler/ghc.mk:653: recipe for target 'compiler/stage2/dll-split.stamp' 
failed
make[1]: *** [compiler/stage2/dll-split.stamp] Segmentation fault

There have been a number of tickets related to dll-split on armhf. I have verified that the ld.gold linker is being invoked. The fixes introduced in #9873 (closed) also seem to be present.

More investigation needed.

Trac metadata

Trac field	Value
Version	7.10.1
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture

mentioned in issue #10234 (closed)

changed milestone to %7.10.2

changed weight to 5

added Tbug Trac import labels

dll-split crashing usually just means the stage2 compiler is broken somehow, it may be good to start by running the test suite.

@rwbarton : With this failure to build dll-split the stage2 compiler isn't complete, so how do I run the test suite?

If dll-split was built then the stage2 compiler has also been built, since dll-split is built with the stage2 compiler. It's just the dll-split check and perhaps some libraries that did not get built yet, but you can run the test suite without those.

A checkout of the ghc-7.10 branch (commit f6c690ba) with these cherry-picked commits:

5d5abdca31     llvmGen: move to LLVM 3.6 exclusively
56fbc18e13     Commit missing T10148 files and ignore the built executable.

and commit that fixes some configure.ac, aclocal.m4 and distrib/configure.ac.in issues does not suffer from this ddl-split issue.

This is also reminiscent of #9920 (closed) where the fix was to either patch the llvm-3.5 sources or move to llvm-3.6.

Its seems to that @bgamari is is using llvm-3.5.1 where I have been using the llvm-3.5 package from Debian (which seems to be llvm-3.5.0 with about 30 patches applied). Since ticket #9920 (closed) actually has the arm-lower-tail-calls.patch patch as an attachment, that patch could probably be added the Debian's llvm package, but that doesn't solve the problem on non-Debian systems.

What we need now is a way to detect this llvm bug at configure time. I'm working on that.

closed

The patch to detect this at configure time is in #10329 (closed) and D856 .

Trac metadata

Trac field	Value
Resolution	Unresolved → ResolvedDuplicate

added ghc build failure label

added Pnormal label