On 16.06.2021 15:08, Andrew Cooper wrote: > On 16/06/2021 07:44, Jan Beulich wrote: >> On 15.06.2021 18:19, Andrew Cooper wrote: >>> --- a/tools/tests/x86_emulator/Makefile >>> +++ b/tools/tests/x86_emulator/Makefile >>> @@ -7,10 +7,6 @@ TARGET := test_x86_emulator >>> .PHONY: all >>> all: >>> >>> -.PHONY: run >>> -run: $(TARGET) >>> - ./$(TARGET) >>> - >>> # Add libx86 to the build >>> vpath %.c $(XEN_ROOT)/xen/lib/x86 >>> >> This is not only incomplete, but actively (specifically here for my >> own frequent using of it, but other tests I do run occasionally as >> well, and then also that same way) harming things as long as you >> don't introduce an alternative way. Note the top-level Makefile >> making use of these rules, and note also the run32 companion here. > > Honestly, this makefile is borderline impossible to follow. I failed to > make the install runes work, which is part of why I deferred the > unit-like tests for now.
Well, yes, it's not pretty, but I lack ideas for a clear improvement. > But I'm taking this as a strong preference to keep the run target? Yes, and not just here. Any test that can be run directly in the build tree would imo better have such a target, so re-building and running it can be invoked with a single make invocation. Having to re-build the entire tools/ in order to then (re-)run one of these tests is, in particular, way too much latency. Jan
