[cfe-dev] A problem with forEachConstructorInitializer

Pedro Delgado Perez pedro.delgado at uca.es
Fri Sep 27 00:47:58 PDT 2013


 
Yes, for instance, I ported back the matcher parmVarDecl and it seemed to work fine (although right now is not working in the matcher

recordDecl(hasMethod(has(parmVarDecl(hasName("x")))))

and I don't know why).

However, it seems not possible to port back the matchers ctorInitializer or forEachConstructorInitializer as some files need to be changed... and this is a problem to me.
1. Is there a site where I can find the steps to correctly update clang to the version 3.4?
2. Has anyone been able to fix at least these matchers?

Thanks,

Pedro.


El día 26 sep 2013 23:54, Gabor Kozar <kozargabor at gmail.com> escribió:
> In general you can't just go around updating random Clang files in hopes of getting some select new features. You should seriously consider either moving to the trunk version, or just porting back the required features. Sometimes though it's just not possible, because a feature e.g. relies on structural support from the system - for instance, I was looking at trying to port back equalsBoundNode() to Clang 3.3, but it's not possible.
>  
> So your best bet is probably looking at the implementation of the new Clang 3.4 features you want to use with your older version, and try to figure out if it can be ported back... Matchers usually can be.
>  
> 
> -- 
> Gábor Kozár -- ShdNx
> kozargabor at gmail.com
>  
>  
>  
> On Thu, Sep 26, 2013, at 17:01, Pedro Delgado Perez wrote:
>> Your issue is most confusing. You will need to show us the matcher code you're trying to get working.
>> 
> I've been looking for the 'forEachConstructorInitializer' and it wasn't in ASTMatchers.h. I've been trying to replace these files(macros, internal, matchfinder...) for the ones in the svn, but it didn't work. 
> 
> I don't want you to be thinking on the matcher, but my matcher, trying to bind every constructor initializer for fields in the class, is:
> 
> constructorDecl(
> 
>                isDefinition(), 
>                anyOf(
>                   hasAnyConstructorInitializer(isWritten()), has(compoundStmt(has(stmt())))
>                ),
>                forEachConstructorInitializer(forField(decl().bind("CI")))      
>          );
> I have tried to change 'forEachConstructorInitializer' for 'forEach(ctorInitializer)', but the system can't identify ctorInitializer neither! :S
> 
> As for your build issues: try doing a clean rebuild, i.e. reconfigure (or run cmake again), and then make. I have never used ninja, so no idea if the issue could be on that front. Also make sure that you've not accidentally corrupted your LLVM/Clang checkout... But I hope you already tried these.
> 
I don't know why, but if I run ninja with "sudo", it works fine now. 

El día 26 sep 2013 14:28, Gabor Kozar <kozargabor at gmail.com> escribió:
Replying on cfe-dev.
 
Your issue is most confusing. You will need to show us the matcher code you're trying to get working.
 
By the way, you do not need to place your own matchers in ASTMatchers.h, you can create an own header file, like so:
 
#include "clang/ASTMatchers/ASTMatchersMacros.h"
#include "clang/ASTMatchers/ASTMatchersInternal.h"
 
namespace clang {
namespace ast_matchers {
 
// your matchers here
 
} // end namespace ast_matchers
} // end namespace clang
 
Do not forget the header guard of course and anything else you may need.
 
As for your build issues: try doing a clean rebuild, i.e. reconfigure (or run cmake again), and then make. I have never used ninja, so no idea if the issue could be on that front. Also make sure that you've not accidentally corrupted your LLVM/Clang checkout... But I hope you already tried these.
 

-- 
Gábor Kozár -- ShdNx
kozargabor at gmail.com
 
 
 
On Thu, Sep 26, 2013, at 11:29, Pedro Delgado Perez wrote:
Hi Gábor,

Sorry to write you directly, but I don't know why I can't post messages in cfe-dev...

I get through you because, some months ago, you told me what could I do when a matcher in http://clang.llvm.org/docs/LibASTMatchersReference.html [ http://clang.llvm.org/docs/LibASTMatchersReference.html ]

was no present in my ASTMatchers.h.
Now, I'm trying to integrate the matcher forEachConstructorInitializer in

http://llvm.org/svn/llvm-project/cfe/trunk/include/clang/ASTMatchers/ASTMatchers.h [ http://llvm.org/svn/llvm-project/cfe/trunk/include/clang/ASTMatchers/ASTMatchers.h ]
Firstly, I have included it in my ASTMatchers, but I don't know why the system is not recognizing this new matcher. So I have put it in my own source file. Now, the system gives me the next problem and I would like to know how could I solve this:

CTD_operator.cpp:42:29: error: calling a private constructor of class 'clang::ast_matchers::internal::BoundNodesTreeBuilder'

            BoundNodesTreeBuilder InitBuilder(*Builder);
                                  ^
/usr/local/include/clang/ASTMatchers/ASTMatchersInternal.h:178:3: note: declared private here
 BoundNodesTreeBuilder(const BoundNodesTreeBuilder &) LLVM_DELETED_FUNCTION;
 ^
CTD_operator.cpp:45:25: error: no viable conversion from 'clang::ast_matchers::internal::BoundNodesTreeBuilder' to 'const clang::ast_matchers::internal::BoundNodesTree'
              Result.addMatch(InitBuilder);
                              ^~~~~~~~~~~
/usr/local/include/clang/ASTMatchers/ASTMatchersInternal.h:113:7: note: candidate constructor (the implicit copy constructor) not viable: no known conversion from
     'clang::ast_matchers::internal::BoundNodesTreeBuilder' to 'const clang::ast_matchers::internal::BoundNodesTree &' for 1st argument
class BoundNodesTree {
     ^
/usr/local/include/clang/ASTMatchers/ASTMatchersInternal.h:172:39: note: passing argument to parameter 'Bindings' here
 void addMatch(const BoundNodesTree& Bindings);
                                     ^
CTD_operator.cpp:48:14: error: 'operator=' is a private member of 'clang::ast_matchers::internal::BoundNodesTreeBuilder'
          *Builder = Result;
          ~~~~~~~~ ^ ~~~~~~
/usr/local/include/clang/ASTMatchers/ASTMatchersInternal.h:179:8: note: declared private here
 void operator=(const BoundNodesTreeBuilder &) LLVM_DELETED_FUNCTION;
      ^
 
Should I also change ASTMatchersInternal.h?? Do you recommend me to replace these header files for the last version?

How can I achieve that the system is aware of the changes I introduce in these files? I have a Makefile and only write "make" from the shell... Since two days ago, I was using ninja, but I don't know why, suddenly, start giving this fail:

pedro at hal ~/clang-llvm/build $ ninja

[6/745] Building CXX object lib/Target/Hexagon/CMakeFiles/LLVMHexagonCodeGen.dir/HexagonRemoveSZExtArgs.cpp.o

FAILED: /usr/local/bin/clang++  -D_DEBUG -D_GNU_SOURCE -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS -fPIC -fvisibility-inlines-hidden -Wall -W -Wno-unused-parameter -Wwrite-strings -Wmissing-field-initializers -pedantic -Wno-long-long -Wcovered-switch-default -Wnon-virtual-dtor -fcolor-diagnostics -fno-rtti -Ilib/Target/Hexagon -I/home/pedro/clang-llvm/llvm/lib/Target/Hexagon -Iinclude -I/home/pedro/clang-llvm/llvm/include   -fno-exceptions -fno-exceptions -MMD -MT lib/Target/Hexagon/CMakeFiles/LLVMHexagonCodeGen.dir/HexagonRemoveSZExtArgs.cpp.o -MF "lib/Target/Hexagon/CMakeFiles/LLVMHexagonCodeGen.dir/HexagonRemoveSZExtArgs.cpp.o.d" -o lib/Target/Hexagon/CMakeFiles/LLVMHexagonCodeGen.dir/HexagonRemoveSZExtArgs.cpp.o -c /home/pedro/clang-llvm/llvm/lib/Target/Hexagon/HexagonRemoveSZExtArgs.cpp
error: unable to open output file 'lib/Target/Hexagon/CMakeFiles/LLVMHexagonCodeGen.dir/HexagonRemoveSZExtArgs.cpp.o': 'Error opening output file 'lib/Target/Hexagon/CMakeFiles/LLVMHexagonCodeGen.dir/HexagonRemoveSZExtArgs.cpp.o''
1 error generated.
[6/745] Linking CXX static library lib/libLLVMInstrumentation.a
ninja: build stopped: subcommand failed.
 
 
I don't know what is "HexagonRemoveSZExtArgs.cpp.o" :S
Thanks,

Pedro.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20130927/d273efa6/attachment.html>


More information about the cfe-dev mailing list