Paul D. Smith <[EMAIL PROTECTED]> writes: > Anyway, the idea was certainly performance: if a directory doesn't exist > and there are no rules in your makefile to create it (that is, there are > no targets that allow make to create that directory), then in make's > model it can never exist, so removing it from the search path is a > performance improvement. For directories that are often not there > (think the CVS or RCS subdirectories, etc. which not only aren't there > much of the time, but are associated with "match anything" rules) this > can be significant.
I understand that. But why would make consider one-level non-existence acceptable? E.g., if /tmp/out exists and /tmp/out/bin does not then /tmp/out/bin/foo can be built. If /tmp/out does not exist then it cannot. Go figure... > However, this check was causing more problems than it was worth (to me) > and I've already removed it from the next release. It is still there in anon-cvs unless we are talking about different places. -boris
signature.asc
Description: Digital signature
_______________________________________________ Bug-make mailing list [EMAIL PROTECTED] http://lists.gnu.org/mailman/listinfo/bug-make
