On 05/10/2016 02:33 AM, Rob Landley wrote: > Sorry, the distraction was actually that when I tried your test I had > that darn "./ps -A doesn't show all processes, but a make clean fixes > it" thing again and I snapshotted the generated/ directory and compared > them and it makes NO sense. (I suspect I overwrite some generated files > with a later build. I still don't have a reproduction sequence.) > > Attempting to diff the "objdump -d generated/obj/ps.o" output was... > long and uninformative. > > Distracted me from the topic at hand a bit. 99% sure this is a build > dependency issue and not a code issue though. It _is_ reliably fixed by > "make clean", I just... ???
Reproduction sequence: make clean make top make ps $ ./ps -A PID TTY TIME CMD 21087 pts/81 00:00:00 ps 30665 pts/81 00:00:00 bash The bug is building other standalone commands in the same file. And the problem is the dependencies don't rebuild generated/obj/ps.o because it's newer than generated/flags.h but _not_ newer than generated/config.h but since generated/config.h is regenerated every time if I make stuff depend on that we essentially always rebuild everything. Hmmm, I can make generated/config.h depend on .config and thus only selectively rebuild it? I should go poke at that but it's week 2 of my trip to tokyo and as usual, I have a stomach bug. (Microbiomes and incubation time: still a thing.) > Rob Still Rob, but urp-y. _______________________________________________ Toybox mailing list [email protected] http://lists.landley.net/listinfo.cgi/toybox-landley.net
