Please keep the mailing list on the CC-line... On Thu, Jul 26, 2012 at 11:42 PM, NAKAMURA Takumi <[email protected]>wrote:
> Chandler, > > I think it would not be the right way that preceding(dependent) rules > and actions satisfy generating headers. > > My intention is; Each of modules should have rules for itself to > generate required headers. > I think we fundamentally disagree here. What rules generate which AST headers is an implementation detail of the AST library (for instance). I do not want that information leaked to users of the AST library, they should just depend on the AST library and be done with it. > I am working for; > > - LLVMBuild-izing clang > Why? I think this would be a step in the wrong direction. Currently Clang has no need of this -- adding it would add complexity and overhead to the understanding of the build system for no tangible gain. > - Cutting inter-libs deps I think we need to preserve proper layering and *simple* deps that are easily understood. I also think we need the system to be simple, easy to understand, and easy for developers to get right. What you're proposing doesn't provide enough encapsulation for this. With them, "make -j100" reaches loadavg to 100.0 :p > I have no such load problems today. I think there must be some other cause that only happens to be solved by the changes you are making, and frankly, I don't think the changes justify the cost. > I shall propose patches when I am ready. Please discuss then. > No, we should discuss here. ;] You're already committing changes that violate the encapsulation we previously had, and otherwise shifting the cmake build in a direction that we've talked about before, and I've expressed specific objections to. I think we need to resolve the direction issues now. =D -Chandler
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
