rboston628
commented
5 months ago I think the simplest solution is to only run the tests on unix, but run the builds on unix, macOS, and windows.
Open rboston628 opened 5 months ago
rboston628
commented
5 months ago I think the simplest solution is to only run the tests on unix, but run the builds on unix, macOS, and windows.
rboston628
commented
3 months ago I've looked more into Catch. It looks simple enough, and I have done some work figuring out how to integrate it into a makefile build.
Consider this file tree
-|
|-- src/
| |-- model/
| | |-- classA.cpp
| | |-- classB.cpp
| | |-- classA.hpp
| | |-- classB.hpp
| |-- main.cpp
|-- tests/
| |-- test_classA.cpp
| |-- test_classB.cpp
|-- makefileThen the makefile should be
# files to pull headers/source or to wite object code
IDIR=src
SDIR=src
ODIR=obj
# compiler and flags
CC=g++ --std=c++14
CFLAGS=-I$(IDIR) -Wall
## specify the needed files, separated as model and main
_MAIN = main
_MODEL = classA classB
# create path to class header/source
MODELDPS = $(patsubst %, $(IDIR)/model/%.hpp, $(_MODEL))
MODELSRC = $(patsubst %, $(SDIR)/model/%.cpp, $(_MODEL))
MAINSRC = $(patsubst %, $(SDIR)/%.cpp, $(_MAIN))
# object files
MODELOBJ = $(patsubst %, $(ODIR)/model/%.o, $(_MODEL))
MAINOBJ = $(patsubst %, $(ODIR)/%.o, $(_MAIN))
# specify code dependencies
MODELDEP = $(MODELDPS) $(MODELSRC)
MAINDEP = $(MAINSRC) $(MODELDEP)
## Main rule for creating the code base
## it will create model and main object files
code: $(MAINOBJ) $(MODELOBJ)
$(CC) -o $@ $^ $(CFLAGS)
## Rules for each subsection -- only update if their dependencies change
$(MODELOBJ): $(ODIR)/model/%.o: $(SDIR)/model/%.cpp $(MODELDEP)| obj/model/
$(CC) -c -o $@ $< $(CFLAGS)
$(MAINOBJ): $(ODIR)/%.o: $(SDIR)/%.cpp $(MAINDEP)| obj/
$(CC) -c -o $@ $< $(CFLAGS)
## create the tests
TEST = $(patsubst %, ./tests/test_%.cpp, $(_MODEL))
tests: code
$(CC) -o ./tests/tests.o $(MAINOBJ) $(MODELOBJ) $(TEST) $(CFLAGS) -l catch2
./tests/tests.o --success
## this rule creates an obj directory if none exists
obj/:
mkdir -p obj
obj/model/:
mkdir -p obj/model
I raised an issue on CxxTest about failing runs on macOS runners. Given their track record of responding in a timely manner to issues, I expect it will never be fixed.
For reasons unknown to me, all tests on github's macOS runners are failing with the following:
As shown, there are multiple syntax errors inside the python script
cxxtest_parser.pyusingisfor comparison of non-Nonevalues, and then a failure to find thecxxtest/TestListener.hheader file.Given several other issues that have occurred with CxxTest, I should consider a new testing framework.