David Bustos wrote: > Quoth Liane Praza on Tue, Oct 14, 2008 at 09:58:26PM -0700: >> David Bustos wrote: >>> Quoth Liane Praza on Thu, Oct 09, 2008 at 08:49:01PM -0700: >>>>>>>> http://cr.opensolaris.org/~lianep/webrev-20080822/ >>>>>>> lib/libscf/common/scf_tmpl.c > ... >>>>>>> 4162,68,72: Is there a reason to return SCF_ERROR_CONSTRAINT_VIOLATED >>>>>>> instead of SCF_ERROR_TEMPLATE_INVALID? > ... >>> Ah, yes, the developer could be calling a count function on an int >>> property. If min > max, though, doesn't that necessarily mean the >>> template is invalid? I.e., that the developer delivering the manifest >>> made an error, rather than the developer calling the function? And for >>> the int functions, if the value doesn't fit in an int64_t, doesn't that >>> also mean that the developer delivering the manifest made an error, >>> rather than the developer calling the function? >> But in the count representation, -1 > 1, which isn't true in the int >> representation. Template isn't invalid. > > I presume you mean the case where the string contains "-1,1". Won't > strtoull() fail on the '-' before we reach the min > max comparison?
It doesn't fail in the test I just wrote. > And what about the int out-of-range cases? The checks for errno are correct, as the number may be too large to represent in an int, but not a count. I could change the check for min > max in the int function, but it'll probably invalidate some of our tests. Think it's important enough to make changes both in the tests and the code? If so, I can. >>>>>>> cmd/svc/milestone/restarter.xml >>>>>>> 80: Isn't restarter created by librestart? [...] > Ok. Please add a comment. Done. liane
