Alexander Leidinger <alexan...@leidinger.net> writes:You change from "no fork+exec if the file doesn't exist" (due to "if" and "[" being shell-builtins) to "always fork+exec". On fast machines surely not an issue, on slow ones, it may make a difference (I have an old amd64 machine at an ISP which takes days to do a buildworld with -j2 due to not much memory, only 2 cores, old HDs, and other stuff going on in parallel).Have you measured this? Because the whole point of clean_dep() is that
No, I haven't measured, for this reason I used "may".
the file it looks for nearly always exists. It's the grep we're not sure of. So checking if the file exists is nearly always a waste.
Thanks for clarifying that.
While the .depend.* namespace is surely controlled by us, would it make sense to change the glob to ".{o,pico}" instead of ".*o" instead to prevent unexpected surprises in the future?Our sh does not support the {} syntax. Besides, what would it change? What else would match .*o but not .{o,pico}?
My point here is, that making this more specific (we can also list the files explicitely), we _maybe_ can prevent foot-shooting in the _future_, in case there is something else added. I've seen enough cases where wildcarded removals caused harm after a little change in some other place. I have no doubts that you checked that this _currently_ is working fine. Based upon my experience in troubleshooting issues with "lost files" after an innocent change somewhere else, I prefer to at least ask about it.
Bye, Alexander. -- http://www.Leidinger.net alexan...@leidinger.net: PGP 0x8F31830F9F2772BF http://www.FreeBSD.org netch...@freebsd.org : PGP 0x8F31830F9F2772BF
pgpVgO_3zVeBo.pgp
Description: Digitale PGP-Signatur
