[llvm] r198441 - For disassembly when adding a symbolic operand that is a C++
Kevin Enderby
enderby at apple.com
Fri Jan 3 16:08:03 PST 2014
Also note, this is simply adding a comment to the disassembled output like this:
% otool -tV /Volumes/SandBox/build-llvm/Debug+Asserts/bin/llvm-mc
...
0000000100001034 callq __ZNK4llvm6Target20createMCDisassemblerERKNS_15MCSubtargetInfoE ## llvm::Target::createMCDisassembler(llvm::MCSubtargetInfo const&) const
And it is just an aid to the reader.
We could hide the demangling as a new ReferenceType, say LLVMDisassembler_ReferenceType_DeMangled_Name, and return it indirectly via ReferenceName to the SymbolLookUp() call back. That way if the user of the C API wants to do a demangling of the name it could be done there. Happy to implement this in this way and that leaves it to the user of the C API.
My thoughts,
Kev
On Jan 3, 2014, at 3:49 PM, Reid Kleckner <rnk at google.com> wrote:
> OTOH a disassembler is not a compiler. It's just a debugging aid. While it's conceptually nice that llvm-c-test --disassemble produces the same outputs on all platforms, it's not as important as being able to do perfect cross compilation.
>
> Is there a way to structure this so that the symbolization is delegated to the user of the C API, or is that too disruptive?
>
>
> On Fri, Jan 3, 2014 at 3:37 PM, David Majnemer <david.majnemer at gmail.com> wrote:
> Unless I am mistaken, the correct thing to do would be to provide an implementation of symbol demangling within LLVM itself.
> This way we could demangle Itanium-style symbols without relying on ::abi::__cxa_demangle and MSVC-style symbols without relying on ::UnDecorateSymbolName.
>
> Relying on the host's ABI details to help disassemble code for a target seems like the wrong approach.
>
>
> On Fri, Jan 3, 2014 at 12:06 PM, Howard Hinnant <hhinnant at apple.com> wrote:
> abi::__cxa_demangle and <cxxabi.h> are part of the Itanium ABI. MS doesn't support this unofficial standard. Sorry, I don't know how to demangle names on Windows. I believe that their type_info.name() returns an demangled name. But I don't know that this info helps.
>
> Howard
>
> On Jan 3, 2014, at 3:02 PM, Kevin Enderby <enderby at apple.com> wrote:
>
> > Hi Reid,
> >
> > Sorry about that. I'll revert my change.
> >
> > Not sure the best way to get this done if an ifdef is not acceptable. Maybe Howard or others will have better suggestions.
> >
> > It does not have any tests as this is used by darwin's otool(1) that links with the llvm disassembler. We have had this problem with this part of the disassembler since day one. So I'm not sure the best way to address that either.
> >
> > Kev
> >
> > On Jan 3, 2014, at 11:54 AM, Reid Kleckner <rnk at google.com> wrote:
> >
> >> Microsoft doesn't ship a <cxxabi.h> header:
> >> http://lab.llvm.org:8011/builders/lld-x86_64-win7/builds/5504/steps/build_Lld/logs/stdio
> >> MCExternalSymbolizer.cpp(16): fatal error C1083: Cannot open include file: 'cxxabi.h': No such file or directory
> >>
> >> I'm not familiar with this code, but it seems bad to make the result of LLVMDisasmInstruction dependent on the target platform, so throwing this behind a quick ifdef isn't necessarily right either.
> >>
> >> This also doesn't have any tests.
> >>
> >>
> >> On Fri, Jan 3, 2014 at 11:33 AM, Kevin Enderby <enderby at apple.com> wrote:
> >> Author: enderby
> >> Date: Fri Jan 3 13:33:09 2014
> >> New Revision: 198441
> >>
> >> URL: http://llvm.org/viewvc/llvm-project?rev=198441&view=rev
> >> Log:
> >> For disassembly when adding a symbolic operand that is a C++
> >> symbol name, also put the human readable name in a comment.
> >>
> >> Also fix a bug in LLVMDisasmInstruction() that was not flushing
> >> the raw_svector_ostream for the disassembled instruction string
> >> before copying it to the output buffer that was causing truncation
> >> of the output.
> >>
> >> rdar://10173828
> >>
> >> Modified:
> >> llvm/trunk/lib/MC/MCDisassembler/Disassembler.cpp
> >> llvm/trunk/lib/MC/MCExternalSymbolizer.cpp
> >>
> >> Modified: llvm/trunk/lib/MC/MCDisassembler/Disassembler.cpp
> >> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCDisassembler/Disassembler.cpp?rev=198441&r1=198440&r2=198441&view=diff
> >> ==============================================================================
> >> --- llvm/trunk/lib/MC/MCDisassembler/Disassembler.cpp (original)
> >> +++ llvm/trunk/lib/MC/MCDisassembler/Disassembler.cpp Fri Jan 3 13:33:09 2014
> >> @@ -298,6 +298,7 @@ size_t LLVMDisasmInstruction(LLVMDisasmC
> >> emitLatency(DC, Inst);
> >>
> >> emitComments(DC, FormattedOS);
> >> + OS.flush();
> >>
> >> assert(OutStringSize != 0 && "Output buffer cannot be zero size");
> >> size_t OutputSize = std::min(OutStringSize-1, InsnStr.size());
> >>
> >> Modified: llvm/trunk/lib/MC/MCExternalSymbolizer.cpp
> >> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCExternalSymbolizer.cpp?rev=198441&r1=198440&r2=198441&view=diff
> >> ==============================================================================
> >> --- llvm/trunk/lib/MC/MCExternalSymbolizer.cpp (original)
> >> +++ llvm/trunk/lib/MC/MCExternalSymbolizer.cpp Fri Jan 3 13:33:09 2014
> >> @@ -13,6 +13,7 @@
> >> #include "llvm/MC/MCInst.h"
> >> #include "llvm/Support/raw_ostream.h"
> >> #include <cstring>
> >> +#include <cxxabi.h>
> >>
> >> using namespace llvm;
> >>
> >> @@ -56,6 +57,14 @@ bool MCExternalSymbolizer::tryAddingSymb
> >> if (Name) {
> >> SymbolicOp.AddSymbol.Name = Name;
> >> SymbolicOp.AddSymbol.Present = true;
> >> + // If Name is a C++ symbol name put the human readable name in a comment.
> >> + if (strncmp(Name, "__Z", 3) == 0) {
> >> + char *demangled = abi::__cxa_demangle(Name + 1, 0, 0, 0);
> >> + if (demangled) {
> >> + cStream << demangled;
> >> + free(demangled);
> >> + }
> >> + }
> >> }
> >> // For branches always create an MCExpr so it gets printed as hex address.
> >> else if (IsBranch) {
> >>
> >>
> >> _______________________________________________
> >> llvm-commits mailing list
> >> llvm-commits at cs.uiuc.edu
> >> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> >>
> >
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140103/ff493e96/attachment.html>
More information about the llvm-commits
mailing list