In http://reviews.llvm.org/D7639#136280, @xazax.hun wrote:
> In http://reviews.llvm.org/D7639#136129, @LegalizeAdulthood wrote: > > > If you can show me how to lex a SourceRange without getting that error, I'm > > happy to change it. I tried the method shown in the linked diff and I get > > the same error. > > > As far as I know, the point is that you can not do that without creating a > null terminated string, however you do not need to do that. What you can do > insead: you can lex the whole file buffer starting from the beginning of a > declaration and you can make sure that you finish lexing on the end of the > declaration (e.g.: end lexing when you match the closing paren of the > declaration). This way you might have a bit more complicated lexing logic but > you avoid heap allocation. So the point is: you construct the lexer using the > whole file buffer starting from the right position and not relying on the > source range at all to finish lexing. Relex the entire file just to avoid a heap allocation? That seems a bit excessive. Do we have any measurements on real code bases that show this to be the better approach? I don't want to optimize like this without real data to show that it is worthwhile. http://reviews.llvm.org/D7639 EMAIL PREFERENCES http://reviews.llvm.org/settings/panel/emailpreferences/ _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
