On 01/02/2019 22:40, Luc Van Oostenryck wrote: > On Fri, Feb 01, 2019 at 09:01:20PM +0000, Ramsay Jones wrote: >> >> I suspect that the Makefile sparse target is not easy to use in a CI >> job, since the 'sparse' program (via cgcc -no-compile) does not exit >> with a non-zero value, even when issuing errors and warnings. > > ... > >> We can change that by passing '-Wsparse-error' to 'sparse': >> >> $ make SPARSE_FLAGS=-Wsparse-error change-table.sp >> SP change-table.c >> change-table.h:53:24: error: dubious one-bit signed bitfield >> change-table.h:54:25: error: dubious one-bit signed bitfield >> change-table.h:55:25: error: dubious one-bit signed bitfield >> change-table.h:56:26: error: dubious one-bit signed bitfield >> Makefile:2729: recipe for target 'change-table.sp' failed >> make: *** [change-table.sp] Error 1 >> $ echo $? >> 2 >> $ >> >> Note that '-Wsparse-error' not only returns a non-zero exit code (1), but >> it also changes a 'warning' into an 'error' (see above): > > Yes, I know :( > The fact that, by default, sparse doesn't fail on errors is wanted > (otherwise it would break the kernel compile). But that the only way > to return an error is to use -Wsparse-error (which is supposed to > replace GCC's -Werror) is a real problem. Given that I only use sparse as a checker, I actually don't mind the current behaviour. That would be different if I was using sparsec/sparsei etc., as a compiler, of course! ;-) [If I were to suggest any change at all to -Wsparse-error it would be: do not change the 'warning' to an 'error' (yes, the actual text label of the message), exit(1) if any errors or warnings, but *if* only warnings have been issued, then print an "treating warnings as errors [-Wsparse-error]" message.] ATB, Ramsay Jones
