Convert pcl::int_t to std::int_t

[kunaltyagi]

Reduces another dependency on boost satisfied by C++14

[taketwo]

Thanks, apart from missing deprecation, LGTM.

[taketwo]

There seem to be a problem with my suggestion, we get warnings of this form:

/__w/1/s/common/include/pcl/Vertices.h:19:17: warning: 'using uint32_t = uint32_t' is deprecated: use std::uint32_t instead [-Wdeprecated-declarations]
     std::vector<uint32_t> vertices;

Not sure what is the best course of action. Do we want to force people to always use std:: prefix? If so, we should also use inside the library, right?

[kunaltyagi]

we should also use inside the library, right

Yes. I'll fix those

[kunaltyagi]

Fixing by using the following script (not exactly, modified)

for type in uint8_t int8_t .... int_fast16_t; do
  for file in grep "[ <(]$type" * -nrl; do
    # had an issue with capture gp in sed
    sed -i "s/ $type/ std::$type/g" $file
    sed -i "s/<$type/[std::$type/g" $file
    sed -i "s/($type/(std::$type/g" $file
    sed -i 's/std::std::/std::/g' $file
  done
done

The change in the last 2 commit will be large. I should have used clang-tidy, but I realized too late.

Since I used sed, some changes will affect whitespace. I haven't checked the diff to see where and will rectify them later

[kunaltyagi]

PTAL. No more deprecation warning regarding pcl::u*int_.*_t

[taketwo]

Thanks, looks good! One thing on my mind though is the deprecation message. As it stands, we get

warning: ‘using uint8_t = uint8_t’ is deprecated: use std::uint8_t instead [-Wdeprecated-declarations]

which is confusing. What do you think about deviating from our usual template and stating explicitly instead of what std::uint8_t should be used? Something like:

warning: ‘using uint8_t = uint8_t’ is deprecated: use std::uint8_t instead of pcl::uint8_t [-Wdeprecated-declarations]

[kunaltyagi]

Sure. That's a good idea actually. Users who use using namespace pcl will see this weird message since the deprecation is in the namespace

[SunBlack]

Just to mention: I'm getting now an compile issue since this PR:

[ 27%] Building CXX object surface/CMakeFiles/pcl_surface.dir/src/gp3.cpp.o
/media/sf_pcl/gpu/surface/src/internal.h(47): error: qualified name is not allowed

/media/sf_pcl/gpu/surface/src/internal.h(52): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/device.h(74): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(69): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(69): error: class "pcl::device::Static<<error-constant>>" has no member "check"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(265): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(317): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(323): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(337): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(349): error: no instance of overloaded function "thrust::sort_by_key" matches the argument list
            argument types are: (thrust::device_ptr<<error-type>>, <error-type>, thrust::device_ptr<int>)

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(362): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(374): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(375): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(591): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(616): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(692): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(734): error: namespace "std" has no member "uint64_type"

/media/sf_pcl/gpu/surface/src/cuda/convex_hull.cu(735): error: namespace "std" has no member "uint64_type"

18 errors detected in the compilation of "/tmp/tmpxft_00005dbe_00000000-15_convex_hull.compute_75.cpp1.ii".
CMake Error at pcl_gpu_surface_generated_convex_hull.cu.o.RelWithDebInfo.cmake:279 (message):
  Error generating file

[kunaltyagi]

I see. I probably missed the gpu module compilation (along with CI) since I'm using clang not gcc. It might just be a missing header. I'll send a patch ASAP. Sorry

[SunBlack]

I'm using clang6 to build the PCL ;-). I have already fixed it locally, just waiting until build is fully completed, to be sure there are no further compile issues.

It seems you didn't used best search & replace pattern, as this doesn't look as valid code^^:

pcl/gpu/surface/src/convex_hull.cpp

Line 77 in 72a0c4f

pcl::gpu::Static<sizeof(pcl::device::std::uint64_type) == 8>::check();

[kunaltyagi]

didn't used best search & replace pattern

Yeah. I used uint64_t since my sed threw up when I tried to incorporate \b. I thought I'd catch all compilation issues but apparently not. All the errors are in gpu with uint64_type and that further led to them passing through (grep '\b::u*int[0-9]*_t[a-zA-Z]\+\b' * -nr)

I'm using clang6 to build the PCL ;-). I have already fixed it locally

Oh, thanks a lot. Really sorry. I have clang-9 and gcc-9 and cuda doesn't like that. I switched to cuda-gcc (nvcc + gcc-8) and just started on the compilation issues but got side-tracked by:

Edited out platform-specific troubles.

[SunBlack]

Yeah. I used uint64_t since my sed threw up when I tried to incorporate \b.

Oh you are really using sed? I'm using it on our build server, but I don't really like it. You don't get a hint, that sed e.g. doesn't know \d, so I had to use [0-9] instead. But perl on command line isn't much better (in case you have multiline regular expressions). Don't want to know how much hours I spend in last time for this both tools for an easy task...