On Tue, Oct 20, 2015 at 11:11 PM Sean Silva <chisophu...@gmail.com> wrote:
> On Tue, Oct 20, 2015 at 6:45 AM, Manuel Klimek <kli...@google.com> wrote: > >> On Tue, Oct 20, 2015 at 3:38 PM Brad King <brad.k...@kitware.com> wrote: >> >>> On 10/20/2015 04:38 AM, Manuel Klimek wrote: >>> > On Tue, Oct 20, 2015 at 5:52 AM Sean Silva wrote: >>> >> get cmake to generate clang module map files and add explicit module >>> build steps? >>> > >>> > I have some experience hacking on cmake, and from my experience I think >>> > this shouldn't be too hard to get working (mainly work ;) >>> >>> I agree this shouldn't be too hard on the CMake side. Manuel, please >>> come over to the cmake dev list to raise the design discussion. Then >>> we can guide your implementation work. >> >> >> I think Sean volunteered :) (but please keep me cc'ed if you start >> discussing this on cmake-dev) >> >> >>> The main design challenges >>> I foresee are: >>> >>> 1. Deciding how this behavior should be activated for a project by >>> its code and/or by the user. >>> >>> 2. Selection of the proper set of headers if it is not exactly the set >>> listed in the target for some reason. Might this ever by more >>> granular than a whole library target? >>> >> >> I don't think so. >> Main concerns are: >> 1. we need to be able to say something is a "textual" header; those are >> still needed; we can do that by calling them .inc, or by putting something >> into cmake to specify textual headers (that's what we do in our build >> system) >> 2. for the "slow rollout" case we use per-header submodules; but that's >> more an implementation detail than anything else, I think >> > > What is the "slow rollout" case? > The case where you want to migrate parts of the codebase to modules, while other parts are not touched (and you don't necessarily want to do a full bottom-up modularization). "slow" as opposed to "one big swoop" > > -- Sean Silva > > >> >> >>> >>> 3. Finding the right place during the CMake generation process to add >>> the rules for this. >>> >>> We already detect the Clang compiler version so deciding if it is >>> new enough to support the feature should not be hard. >>> >>> Thanks, >>> -Brad >>> >>>
_______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
