This does not change any behaviour, but only passed these variablesalong through the entire stack of functions leading up toObjFileIsUpToDate. This prepares for the next commit.Signed-off-by: Matthijs Kooijman <matthijs@stdin.nl>

If -debug-level=20 is passed, whenever the cached file is not usable forwhatever reason, a message is displayed. This should help debug cachingproblems.The messages are hardcoded in the source and not put into `constants`,since they are only debug messages.Signed-off-by: Matthijs Kooijman <matthijs@stdin.nl>

This slightly cleans up a function call.Signed-off-by: Matthijs Kooijman <matthijs@stdin.nl>

This renames this function and adds a resultFile parameter. ThisresultFile is used for all timestamp checks against other files (sourcefile, dep file, header files from the dep file), while the objectFilepassed is now only used to check against the contents of the dep file.Both calls to this function still pass the same filename as objectFileand resultFile, so this commit should not change any behavior.Signed-off-by: Matthijs Kooijman <matthijs@stdin.nl>

Previously, the including detection process caching relied on the .dfiles generated by compilation to know what .h files a given sourcefiles depends on. This works correctly, but if a project does notcompile completely, not all .d files are generated, so not all cachedinclude detection results can be used.In practice, this means that if a there is a compilation error in thefirst file that is compiled, include detection will run again for allfiles on the next run. If you have a few errors to solve and a bigproject, this gets annoying quickly.To fix this, the include detection process should generate .d filesitself. At first glance it appears that there is a problematic casewhere the list of included header files changes, the include detectionoverwrites the .d file and then compilation only sees the new list(which was generated later than the .o file was generated). However,since this implies that changes are made to an #include directive in thesource file itself or one of the files that are still included, thisshould be detected normally. There is still a corner case when a file ischanged during the build, but that was already the case.Since include detections uses `-o /dev/null`, the compiler generates aslightly different .d file. During compilation, a file `foo.cpp.d` isgenerated in the output directory starting with:    /path/to/foo.cpp.o: \But when just passing `-MMD` to the preproc recipe, it generates a`foo.d` file in the source directory starting with:  foo.o: \To make these equal, `-MF` must be passed during include detection toset the .d filename, and `-MT` must be passed to set the .o filenameinside the .d file.To enable this feature, platform.txt should be modified by adding ` -MMD-MF {dep_file} -MT {object_file}` to `preproc.macros.flags` (or`recipe.preproc.macros`). Without any changes to platform.txt, behaviouris unchanged.To allow this, this adds `{dep_file}` and `{object_file}` variables tothe build properties for the preproc macros recipe. For consistency,`{dep_file}` is also added during normal compilation, though it is notcurrently used.