CI: spurious error building LLVM on mingw: `ranlib.exe: could not create temporary file whilst writing archive: no more archived files`

[erikdesjardins]

# the specific codegen library differs, sometimes libLLVMSparcCodeGen.a, sometimes libLLVMM68kCodeGen.a, etc.
FAILED: lib/libLLVMSparcCodeGen.a
cmd.exe /C "cd . && "C:\Program Files\CMake\bin\cmake.exe" -E rm -f lib\libLLVMSparcCodeGen.a && D:\a\rust\rust\mingw32\bin\ar.exe qc lib\libLLVMSparcCodeGen.a  lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/DelaySlotFiller.cpp.obj lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/LeonPasses.cpp.obj lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/SparcAsmPrinter.cpp.obj lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/SparcInstrInfo.cpp.obj lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/SparcISelDAGToDAG.cpp.obj lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/SparcISelLowering.cpp.obj lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/SparcFrameLowering.cpp.obj lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/SparcMachineFunctionInfo.cpp.obj lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/SparcRegisterInfo.cpp.obj lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/SparcSubtarget.cpp.obj lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/SparcTargetMachine.cpp.obj lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/SparcMCInstLower.cpp.obj lib/Target/Sparc/CMakeFiles/LLVMSparcCodeGen.dir/SparcTargetObjectFile.cpp.obj && D:\a\rust\rust\mingw32\bin\ranlib.exe lib\libLLVMSparcCodeGen.a && cd ."
D:\a\rust\rust\mingw32\bin\ranlib.exe: could not create temporary file whilst writing archive: no more archived files

This has happened on quite a few PRs recently:

• Rebased: Mark drop calls in landing pads cold instead of noinline #102099 (comment)
• Make dyn*'s value backend type a pointer #107772 (comment)
• Make codegen choose whether to emit overflow checks #107921 (comment)
• Remove missing_tools config #107830 (comment)
• add support of RustyHermit's BSD socket layer #107405 (comment)
• make ty::Ty: Debug not call the Display impl #107084 (comment)
• Rollup of 10 pull requests #108096 (comment)
• Rollup of 4 pull requests #107949 (comment)
• Optimize TyKind::eq. #107717 (comment)
• rustc_metadata: Encode/decode some LazyArrays without an Option #107693 (comment)
• Provide structured suggestion for binding needing type on E0594 #107646 (comment)
• Rollup of 6 pull requests #107463 (comment)
• Rollup of 5 pull requests #107456 (comment)
• Enable CopyProp #107449 (comment)
• emit ConstEquate in TypeRelating<D> #107434 (comment)
• assume MIR types are fully normalized in ascribe_user_type #107197 (comment)
• Set RUST_SAVE_ANALYSIS_CONFIG env variable for run-pass tests #106789 (comment)
• rustc_parse: diagnostics migration, v4 #105670 (comment)
• Add explanatory message for [#must_use] in ops #103761 (comment)
• incremental: migrate diagnostics #100754 (comment)

I've seen it happen on the following builders (all mingw builders I guess):

• dist-i686-mingw
• i686-mingw-1
• i686-mingw-2
• dist-x86_64-mingw
• x86_64-mingw-2

@rustbot label A-spurious A-LLVM O-Windows

[klensy]

Ok, internet knows about this problem, https://osdn.net/projects/mingw/lists/archive/users/2019-January/thread.html#198

https://osdn.net/projects/mingw/lists/archive/users/2019-January/000228.html

the buggy mkstemp() has infected the binutils-2.31.1-1, and
gcc-8.2.0-2 builds.

So this should be solved long ago?

[mati865]

I think it started with #100178

Ok, internet knows about this problem, https://osdn.net/projects/mingw/lists/archive/users/2019-January/thread.html#198

https://osdn.net/projects/mingw/lists/archive/users/2019-January/000228.html

the buggy mkstemp() has infected the binutils-2.31.1-1, and
gcc-8.2.0-2 builds.

So this should be solved long ago?

Could be a new regression...