0x8000-0000 added a comment.

In https://reviews.llvm.org/D49114#1156987, @JonasToth wrote:

> I think some of the logic you have in your check code could be done via 
> matchers. That is usually better for performance, because you analyze less 
> code.


I have considered that approach, but had to abandon it because the logic in 
this check is substractive, not additive: I would not use matchers to find more 
loci to diagnose, but to exculpate loci found by more general rules.

In other words: I want to report as diagnostic _all_ integer literals, except 
some, based on the larger construct they are part of.

In my understanding matchers are effective if you need to search for "X 
followed by Y", "X followed by Z", "X followed by Y". The union of the results 
is the set I want.

In my case, I am searching for "all of X", "X unless preceded by A", "X unless 
preceded by B".  The "unless preceded by" rules don't compose well, as their 
union tends to be the whole set "all of X".


Repository:
  rCTE Clang Tools Extra

https://reviews.llvm.org/D49114



_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to