TEST/APPS: Fixed building rpm. #10358
Conversation
|
/azp run UCX release |
|
Azure Pipelines successfully started running 1 pipeline(s). |
rakhmets
force-pushed
the
topic/fix-buildrpm-w-cuda
branch
from
50ce59c to
cb3e27c
Compare
test/apps/Makefile.am
Outdated
| `test -f '$<' || echo '$(srcdir)/'`$< | ||
| $(AM_V_at)$(am__mv) $(DEPDIR)/test_cuda_get_symbol_address.Tpo \ | ||
| $(DEPDIR)/test_cuda_get_symbol_address.Po | ||
| @am__fastdepCC_TRUE@ $(AM_V_CC)$(NVCC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(test_cuda_get_symbol_address_CPPFLAGS) -Xcompiler "$(CPPFLAGS) $(CFLAGS)" -MT $@ -MD -MF $(DEPDIR)/test_cuda_get_symbol_address.Tpo -c -o $@ `test -f 'test_cuda_get_symbol_address.cu' || echo '$(srcdir)/'`test_cuda_get_symbol_address.cu |
There was a problem hiding this comment.
any way to make this simpler? maybe some kind of intermediate/explicit dependency?
what is the issue being solved here?
There was a problem hiding this comment.
This is the same command line that is autogenerated and used to build objects from .c files. The only difference are: $(NVCC) instead of $(CC), -Xcompiler to pass options not supported by nvcc, and the absence of -MP option.
@am__fastdepCC_TRUE@, @AMDEP_TRUE@@am__fastdepCC_FALSE@, @am__fastdepCC_FALSE@ are used to comment out the unused lines depending on the build type.
There was a problem hiding this comment.
Is there some way to avoid taking the autogenerated command and instead just tweak some automake vars?
There was a problem hiding this comment.
Or at least remove some parts of the command which are not essential?
There was a problem hiding this comment.
I did not find an easier way to use two different command line for different compilers. I can try to simplify the command line a little. However, I do not know any other ways to support different command lines for make/make dist/etc... other than using the same variables to comment out the lines.
There was a problem hiding this comment.
I see, so let's try to simplify it
| @am__fastdepCC_TRUE@ $(AM_V_CC)$(test_cuda_get_symbol_address_COMPILE) -MT $@ -MD -MF $(test_cuda_get_symbol_address_DEPBASE).Tpo -c -o $@ `test -f '$(<F)' || echo '$(srcdir)/'`$(<F) | ||
| @am__fastdepCC_TRUE@ $(AM_V_at)$(am__mv) $(test_cuda_get_symbol_address_DEPBASE).Tpo $(test_cuda_get_symbol_address_DEPBASE).Po | ||
| @AMDEP_TRUE@@am__fastdepCC_FALSE@ $(AM_V_CC)source='$(<F)' object='$@' libtool=no @AMDEPBACKSLASH@ | ||
| @AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ | ||
| @am__fastdepCC_FALSE@ $(AM_V_CC@am__nodep@)$(test_cuda_get_symbol_address_COMPILE) -Xcompiler "-fPIC" -c -o $@ `test -f '$(<F)' || echo '$(srcdir)/'`$(<F) |
There was a problem hiding this comment.
can we simplify this further? for example, do we readlly need am__fastdepCC_TRUE/am__fastdepCC_FALSE ? seems it switches between 2 modes of deps, can we use just one?
There was a problem hiding this comment.
Yes, these variables are replaced by # to comment out lines depending on the configure option (dependency-tracking) and the compiler version.
We need these variables at the beginning of the line to compile the object file in the same way as all other object files.
What?
Fixed the following scenario: