On Jun 13, 2011, at 7:02 PM, Peter Collingbourne wrote: > On Mon, Jun 13, 2011 at 03:19:11PM -0700, Anders Carlsson wrote: >> >> On Jun 13, 2011, at 12:41 PM, Peter Collingbourne wrote: >> >>> Hi, >>> >>> This patch refactors the vtable and VTT builders to provide a >>> backend-neutral interface to the builders in the AST library, which >>> CodeGen uses to build vtables and VTTs in LLVM. No functionality >>> change intended. >> >> While I'm not crazy about this idea, other people (Doug) are in favor of it, >> so I'm reluctantly fine with it :) >> >> That being said, I don't think this patch should be committed in its current >> state. This should be done in incremental steps, >> there's no reason to move both the VTT builder and VTable builder in a >> single commit. > > OK, then. Following your and Eli's suggestions I will try to > break this patch up into the following steps: > > 1) Move GlobalDecl to AST > 2) Make VTTBuilder independent of CodeGen (broken into sub-steps) > 3) Move VTTBuilder to AST > 4) Make VTableBuilder independent of CodeGen (broken into sub-steps) > 5) Move VTableBuilder to AST > > As a start, I am attaching a (simple) patch for 1. OK to commit? >
Sure, looks good! >> Also, the files should be svn cp'ed to preserve history; being able to >> easily go back and see the rationale behind changes is >> crucial to fixing bugs/regressions in this code. > > Hmm, I'll see if I can figure out how to persuade git-svn to do that. If you can't persuade git-svn to do it, you could always just svn cp the files as a separate step using, well, svn :) You could then hack on them further using git-svn. - Anders _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
