[CMake] Ninja fails, Make succeeds...

[David Cole]

On Tue, Jun 26, 2012 at 6:40 PM, Mikael Lyngvig <mikael at lyngvig.org> wrote:
> Ohh, sorry.  Ninja says:
>
> [117/325] Building CXX object Driver/CMakeFiles/Driver.dir/Driver.cpp.obj
> FAILED: c:\Mingw64\bin\c++.exe   -DNDEBUG
> -DBR0_BUILD_ASSERT=BR0_BUILD_ASSERT_DELETE
> -DBR0_BUILD_CHARSET=BR0_BUILD_CHARSET_ASCII
> -DBR0_BUILD_ENDIAN=BR0_BUILD_ENDIAN_LITTLE
> -DBR0_BUILD_MODE=BR0_BUILD_MODE_SHIP
> -DBR0_BUILD_TARGET=BR0_BUILD_TARGET_WINDOWS -Wall -Wno-long-long -pedantic
> -I.. -Ic:/Synology/clang-3.2-WIN64/include    -static-libgcc -MMD -MT
> Driver/CMakeFiles/Driver.dir/Driver.cpp.obj -MF
> "Driver/CMakeFiles/Driver.dir/Driver.cpp.obj.d" -o
> Driver/CMakeFiles/Driver.dir/Driver.cpp.obj -c
> C:/Dropbox/Source/braceless/src/Driver/Driver.cpp
> C:/Dropbox/Source/braceless/src/Driver/Driver.cpp:11:42: fatal error:
> Braceless0/Backend/Backend.hpp: No such file or directory
> compilation terminated.
> [117/325] Building CXX object Driver/CMakeFiles/Driver.dir/Setup.cpp.obj
> ninja: build stopped: subcommand failed.
>
> The problem is that the file Driver.cpp depends on Backend.hpp having been
> created.  But Driver.cpp should not be compiled until the second-last item
> in the list of objects to make ([324/325]).
>
> So, somehow, Ninja picks up the target Driver.obj and builds it way too
> early.  With GNU Make, everything goes according to plan, but with Ninja,
> there's some sort of confusion somewhere.
>
>
> Cheers,
> Mikael
>
> 2012/6/27 David Cole <david.cole at kitware.com>
>>
>> On Tue, Jun 26, 2012 at 6:30 PM, Mikael Lyngvig <mikael at lyngvig.org>
>> wrote:
>> > No, I delete the entire build tree (I build out of the source tree,
>> > btw).
>> >
>> > The procedure is as follows:
>> >
>> >    rem foo contains the source code
>> >    md foo-build
>> >    cd foo-build
>> >    cmake -G Ninja ..\foo
>> >    ninja
>> >    rem fails as described earlier
>>
>>
>> But you didn't really describe the failure.
>>
>> What is the error output? (from cmake or from ninja?)
>>
>>
>> >
>> >     cd ..
>> >     rd /s /q foo-build
>> >     md foo-build
>> >     cd foo-build
>> >     cmake -G "MinGW Makefiles" ..\foo
>> >     make
>> >     rem succeeds as described earlier
>> >
>> > Thank you for your reply.  If you need source, I can send you a zip
>> > archive
>> > as it is an incomplete open source project.
>> >
>> >
>> > Cheers,
>> > Mikael
>> >
>> >
>> > 2012/6/27 David Cole <david.cole at kitware.com>
>> >>
>> >> You can't switch generators in the same build tree without deleting
>> >> absolutely everything and starting over.
>> >>
>> >> Is that what you're trying to do?
>> >>
>> >> If you want a make build tree and a ninja build tree, you'll need two
>> >> separate build trees....
>> >>
>> >>
>> >> HTH,
>> >> David
>> >>
>> >>
>> >> On Tue, Jun 26, 2012 at 4:45 PM, Mikael Lyngvig <mikael at lyngvig.org>
>> >> wrote:
>> >> > Hi,
>> >> >
>> >> > I am a fan of Ninja.  Since I started using it, everything's been
>> >> > built
>> >> > much, much faster than before.  Among other things because I use
>> >> > CMake
>> >> > to
>> >> > publish headers to the binary directory so as to allow me to include
>> >> > the
>> >> > headers with a relative path without having to move all my headers to
>> >> > a
>> >> > central include directory (I want my source files and headers to be
>> >> > in
>> >> > the
>> >> > same directory).  Ninja does this very fast, whereas GNU Make takes
>> >> > forever
>> >> > to publish my header files.
>> >> >
>> >> > However, I have a puzzling problem: Out of the blue, Ninja started
>> >> > failing
>> >> > to build my project.  If I remake the project with 'cmake -G "MinGW
>> >> > Makefiles"', it takes ages to build but it succeeds.  If I do 'cmake
>> >> > -G
>> >> > Ninja', it quickly aborts because it seems that either CMake or Ninja
>> >> > wants
>> >> > to build one of the last targets at a very early point in the build
>> >> > phase.
>> >> >  I have an executable, driver.exe, which is supposed to be built as
>> >> > the
>> >> > very
>> >> > last item.  With Ninja, however, it pops up around 20 percent into
>> >> > the
>> >> > build
>> >> > and the build fails because the dependent libraries are nowhere to be
>> >> > found
>> >> > (they haven't been built yet).  I have tried clearing the binaries
>> >> > directory
>> >> > (rd /s /q foo & md foo) and so on a number of times, but Ninja
>> >> > insists
>> >> > on
>> >> > trying to build this final executable at a very early stage.
>> >> >
>> >> > Any ideas?  I suppose the problem is caused by one of these three
>> >> > things:
>> >> >
>> >> >    1. An error on my part.  Very likely.
>> >> >    2. An error in CMake's generation of Ninja scripts.  Not unlikely
>> >> > due
>> >> > to
>> >> > the still fairly immature support for Ninja.
>> >> >    3. An error in Ninja.  Highly unlikely as Ninja builds tons and
>> >> > tons
>> >> > of
>> >> > projects all over the world each day and nobody seem to have run into
>> >> > this
>> >> > problem before.
>> >> >
>> >> > How do you locate the problem?
>> >> >
>> >> > Thanks in advance!
>> >> >
>> >> >
>> >> > Cheers,
>> >> > Mikael
>> >> > -- Love Thy Frog!
>> >> >
>> >> > --
>> >> >
>> >> > Powered by www.kitware.com
>> >> >
>> >> > Visit other Kitware open-source projects at
>> >> > http://www.kitware.com/opensource/opensource.html
>> >> >
>> >> > Please keep messages on-topic and check the CMake FAQ at:
>> >> > http://www.cmake.org/Wiki/CMake_FAQ
>> >> >
>> >> > Follow this link to subscribe/unsubscribe:
>> >> > http://www.cmake.org/mailman/listinfo/cmake
>> >
>> >
>
>

What does your add_custom_command look like that generates Backend.hpp?

(offline for a while, next response much later... :-)