[llvm-dev] setDataLayout segfault
David Jones via llvm-dev
llvm-dev at lists.llvm.org
Wed Sep 14 10:24:19 PDT 2016
I think a large part of the confusion comes from two conflicting ideas:
1. Things that create modules return a unique_ptr<Module>, which indicates
an intention to make it difficult to have multiple copies of a module
"pointer".
2. However, the LLVM API has several methods that return or require bare
Module* pointers.
The result is that you pretty much have to "smash" the unique_ptr<Module>
to get a raw pointer out to get any work done, which is potentially
dangerous.
In my case, I can create a Module in one of two ways:
a) By reading in bitcode:
ErrorOr<std::unique_ptr<llvm::Module> > eom =
parseBitcodeFile(mb->getMemBufferRef(), context);
if (eom) {
m_Module = std::move(eom.get());
b) By newing one:
std::unique_ptr<Module> upm(new Module(name, context));
m_Module = std::move(upm);
If I want to support a) then I am forced to store a unique_ptr<Module> in
my own code.
But then if I want to pass a Module* into some other LLVM API, I need to:
Module &getModule() const { return *m_Module; }
m_Func = Function::Create(as_type, Function::ExternalLinkage, name,
&getModule());
All of this seems quite stilted. On the one hand, we have a strong effort
to ensure that there are no dangling pointers to Modules, but on the other,
we have to jump through hoops to get values required for arguments to other
API calls, and in doing so, undo the "guarantees" we get from using
unique_ptr in the first place.
Or am I totally missing the point? Is there a better way for me to
organize this code?
On Wed, Sep 14, 2016 at 1:02 PM, Jonathan Roelofs via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
>
>
> On 9/14/16 10:23 AM, Frank Winter via llvm-dev wrote:
>
>> Ok. I can make a copy of the unique_ptr before moving it into the
>> builder's constructor and use the copy later on. It is confusing to
>> require a unique_ptr.
>>
>
> It's (almost?) always a bad idea to make a copy of a std::unique_ptr.
>
>
>> Frank
>>
>> On 09/14/2016 12:11 PM, Frank Winter via llvm-dev wrote:
>>
>>> I am constructing the engine builder in the following way:
>>>
>>> llvm::SMDiagnostic Err;
>>> unique_ptr<Module> Mod = getLazyIRFileModule("f.ll", Err, TheContext);
>>> llvm::EngineBuilder engineBuilder(std::move(Mod));
>>>
>>> However, after moving the pointer to the constructor it is no longer
>>> retrievable from the unique_ptr object.
>>>
>>
> Moving re-assigns ownership of the pointed-to memory to the EngineBuilder.
>
>
>>> Mod.get()->dump(); // this segfaults after the move, but not before
>>>
>>
> That is expected.
>
>
>>> So I conclude that any type of operation on the module is no longer
>>> valid.
>>>
>>
> Correction: any operation via the *moved* unique_ptr is no longer valid.
>
>
>>> Am I constructing the engine builder as it is supposed to?
>>>
>>> Frank
>>>
>>>
>>> On 09/14/2016 10:59 AM, Frank Winter via llvm-dev wrote:
>>>
>>>> I get a segfault with this code when setting the data layout:
>>>>
>>>>
>>>>
>>>> int main(int argc, char** argv)
>>>> {
>>>> llvm::InitializeNativeTarget();
>>>>
>>>> llvm::LLVMContext TheContext;
>>>> unique_ptr<Module> Mod(new Module("A",TheContext));
>>>>
>>>> llvm::EngineBuilder engineBuilder(std::move(Mod));
>>>> std::string mcjit_error;
>>>>
>>>> engineBuilder.setMCPU(llvm::sys::getHostCPUName());
>>>>
>>>> engineBuilder.setEngineKind(llvm::EngineKind::JIT);
>>>> engineBuilder.setOptLevel(llvm::CodeGenOpt::Aggressive);
>>>> engineBuilder.setErrorStr(&mcjit_error);
>>>>
>>>> llvm::TargetOptions targetOptions;
>>>> targetOptions.AllowFPOpFusion = llvm::FPOpFusion::Fast;
>>>> engineBuilder.setTargetOptions( targetOptions );
>>>>
>>>> TargetMachine *targetMachine = engineBuilder.selectTarget();
>>>>
>>>> assert(targetMachine && "failed to create target machine");
>>>>
>>>> std::cout <<
>>>> targetMachine->createDataLayout().getStringRepresentation() << "\n";
>>>>
>>>> Mod.get()->setDataLayout(
>>>> targetMachine->createDataLayout().getStringRepresentation() ); //
>>>> this segfaults
>>>> Mod.get()->setDataLayout( targetMachine->createDataLayout() ); //
>>>> as well as this
>>>> }
>>>>
>>>>
>>>> Backtrace:
>>>>
>>>>
>>>> Using host libthread_db library
>>>> "/lib/x86_64-linux-gnu/libthread_db.so.1".
>>>> e-m:e-i64:64-f80:128-n8:16:32:64-S128
>>>>
>>>> Program received signal SIGSEGV, Segmentation fault.
>>>> 0x00007ffff5f65832 in llvm::SmallVectorTemplateCommon<unsigned char,
>>>> void>::end (this=0x148) at
>>>> /home/fwinter/svn/llvm-3.9/include/llvm/ADT/SmallVector.h:117
>>>> 117 iterator end() { return (iterator)this->EndX; }
>>>> (gdb) bt
>>>> #0 0x00007ffff5f65832 in llvm::SmallVectorTemplateCommon<unsigned
>>>> char, void>::end (this=0x148) at
>>>> /home/fwinter/svn/llvm-3.9/include/llvm/ADT/SmallVector.h:117
>>>> #1 llvm::SmallVectorImpl<unsigned char>::clear (this=0x148) at
>>>> /home/fwinter/svn/llvm-3.9/include/llvm/ADT/SmallVector.h:345
>>>> #2 0x00007ffff5fa13b5 in llvm::DataLayout::clear (this=0x138) at
>>>> /home/fwinter/svn/llvm-3.9/lib/IR/DataLayout.cpp:545
>>>> #3 0x00007ffff5f9f561 in llvm::DataLayout::reset (this=0x138,
>>>> Desc=...) at /home/fwinter/svn/llvm-3.9/lib/IR/DataLayout.cpp:179
>>>> #4 0x00007ffff60b25ca in llvm::Module::setDataLayout (this=0x0,
>>>> Desc=...) at /home/fwinter/svn/llvm-3.9/lib/IR/Module.cpp:377
>>>> #5 0x0000000000406c46 in main (argc=1, argv=0x7fffffffde68) at
>>>> main.cc:95
>>>>
>>>>
>>>> Any idea?
>>>>
>>>> Frank
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> LLVM Developers mailing list
>>>> llvm-dev at lists.llvm.org
>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>>>
>>>
>>> _______________________________________________
>>> LLVM Developers mailing list
>>> llvm-dev at lists.llvm.org
>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>>
>>
>> _______________________________________________
>> LLVM Developers mailing list
>> llvm-dev at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>
>
> --
> Jon Roelofs
> jonathan at codesourcery.com
> CodeSourcery / Mentor Embedded
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160914/2a8e9b84/attachment-0001.html>
More information about the llvm-dev
mailing list