Bug: use of undeclared identifier 'FLT_MAX' missing header

[Benoit Daloze]

Hello,

Under certain conditions which are not especially easy to reproduce,
the build of a test using test (starting at r663 until at least r677) fails with:

[ 11%] Building CXX object vm/vm/test/CMakeFiles/vmtest.dir/testutils.cc.o
/usr/bin/clang++    -Wall -std=c++0x -stdlib=libc++ -I/usr/lib/c++/v1 -O3 -DNDEBUG -I/Users/benoitdaloze/git/mozart2/vm/vm/test/../main -I/Users/benoitdaloze/prefix/build/moz2-release-tmp/vm/vm/test/../main -I/Users/benoitdaloze/prefix/build/moz2-release-tmp/externals/Source/gtest -I/Users/benoitdaloze/prefix/build/moz2-release-tmp/externals/Source/gtest/include    -o CMakeFiles/vmtest.dir/testutils.cc.o -c /Users/benoitdaloze/git/mozart2/vm/vm/test/testutils.cc
In file included from /Users/benoitdaloze/git/mozart2/vm/vm/test/testutils.cc:1:
In file included from /Users/benoitdaloze/git/mozart2/vm/vm/test/testutils.hh:5:
In file included from /Users/benoitdaloze/prefix/build/moz2-release-tmp/externals/Source/gtest/include/gtest/gtest.h:58:
/Users/benoitdaloze/prefix/build/moz2-release-tmp/externals/Source/gtest/include/gtest/internal/gtest-internal.h:384:51: error: use of undeclared identifier 'FLT_MAX'
inline float FloatingPoint<float>::Max() { return FLT_MAX; }
                                                  ^
/Users/benoitdaloze/prefix/build/moz2-release-tmp/externals/Source/gtest/include/gtest/internal/gtest-internal.h:386:53: error: use of undeclared identifier 'DBL_MAX'
inline double FloatingPoint<double>::Max() { return DBL_MAX; }
                                                    ^
2 errors generated.

This happens on a build of Mozart2 (https://github.com/mozart/mozart2) on OS X 10.8.5.
For now our fix is to stay at r662 (in vm/vm/CMakeLists.txt) but obviously this is not optimal.

This is most likely due to usage of FLT_MAX and DBL_MAX in gtest-internal.h but not #include-ing <climits> (only <limits>).
Could you therefore add that missing header in gtest-internal.h?

[Billy Donahue]

FLT_MAX and DBL_MAX are required by C to be defined in <float.h>, which IS included in gtest-internal.h.

http://pubs.opengroup.org/onlinepubs/009695399/basedefs/float.h.html

Can you give a minimal test program that shows the problem? I think your code may be doing something unusual.

http://www.cplusplus.com/reference/cfloat/

--
 
---
You received this message because you are subscribed to the Google Groups "Google C++ Testing Framework" group.
To unsubscribe from this group and stop receiving emails from it, send an email to googletestframe...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Benoit Daloze]

On Sunday, March 2, 2014 7:03:28 AM UTC+1, Billy Donahue wrote:

FLT_MAX and DBL_MAX are required by C to be defined in <float.h>, which IS included in gtest-internal.h.

http://pubs.opengroup.org/onlinepubs/009695399/basedefs/float.h.html

You are right, I must have somewhat confused it with std::numeric_limits from <limits>.
 

Can you give a minimal test program that shows the problem? I think your code may be doing something unusual.

http://www.cplusplus.com/reference/cfloat/

This was due to a bad "#define __FLOAT_H" in our float.hh which turns out to be the same #define as the one of Clang's float.h.

So, definitely our fault, sorry I only guessed that #define problem today.

[Billy Donahue]

Hint: Don't use __ for anything ever. It's reserved.

--