On 27 Oct 2010, at 21:58, Andrew Deason wrote: > On Wed, 27 Oct 2010 20:08:34 +0100 > Simon Wilkinson <[email protected]> wrote: > >> Comments? Brickbats? > > I thought one of the reasons for the current system was that CC &c are > forced to certain values because we need to force CC in at least certain > points for building kernel foo. How does this new system cope?
The logic behind picking CC hasn't been changed (yet). On some platforms (Linux, Darwin) we let the user and ./configure pick which C compiler to use, and don't attempt to second guess them. We're already using this for things like running the code base through the Clang static analyser. On other platforms (Solaris, IRIX, AIX, HPUX) we override configure's choice because "we know best". In theory we should also be overriding the flags we pass so that they match the compiler choice, but it's entirely possible we get this wrong. Cheers, Simon. _______________________________________________ OpenAFS-devel mailing list [email protected] https://lists.openafs.org/mailman/listinfo/openafs-devel
