I think that treating an empty file and a missing file equivalently is dangerous because they are in no way semantically equivalent. It is common practice in software to use an empty file as a lock (like with Side-effects).
If nothing else, this makes debugging csig information difficult because missing files have signatures. Unless someone is extremely opposed. I will create an issue on for this if it turns out not to be a quick patch. V/R, William On Sat, Aug 15, 2015 at 6:40 PM, anatoly techtonik <[email protected]> wrote: > On Thu, Jul 30, 2015 at 8:02 PM, William Blevins <[email protected]> > wrote: > >> Team, >> >> I noticed something peculiar when examining how CSIG's are generated for >> Nodes. In the case that get_content returns an error (IE. file is >> missing), then the signature becomes MD5( '' ). >> >> This seems a bit counter-intuitive. Would it make more sense for >> get_csig() to return None until the node can actually be evaluated because >> otherwise a missing file has the same signature as an empty file, and those >> are two very different things. >> > > It depends on usage scenario. For identification purposes it is better to > use prefixes that will reflect additional Node statuses. This will make it > more generic than just missing file check. The question if it is really > what is needed? > -- > anatoly t. > > _______________________________________________ > Scons-dev mailing list > [email protected] > https://pairlist2.pair.net/mailman/listinfo/scons-dev > >
_______________________________________________ Scons-dev mailing list [email protected] https://pairlist2.pair.net/mailman/listinfo/scons-dev
