On Nov 13, 2012, at 10:34 AM, Eric Christopher <[email protected]> wrote:
> On Tue, Nov 13, 2012 at 10:28 AM, Bill Wendling <[email protected]> wrote: > On Nov 13, 2012, at 10:03 AM, Eric Christopher <[email protected]> wrote: > >> On Mon, Nov 12, 2012 at 4:54 PM, Bill Wendling <[email protected]> wrote: >> Author: void >> Date: Mon Nov 12 18:54:24 2012 >> New Revision: 167799 >> >> URL: http://llvm.org/viewvc/llvm-project?rev=167799&view=rev >> Log: >> Disable accelerator tables when compiling with LTO. >> >> LTO doesn't generate correct accelerator tables. This is due to the general >> lack >> correct of debug info for LTO. Disable it when using LTO. >> <rdar://problem/12401423> >> >> >> Not a huge fan of this patch. Debug Info for LTO in general isn't working >> yet, what's the difference between this and anything else? What kind of >> configuration are you building with that this is occurring? What tool is >> breaking? >> > If we build LLVM in LTO mode and then run dtrace on the executable, it will > eat all of your memory and fill up your hard drive because of bad debug info > resulting from bad accelerator tables. > > > dtrace uses the accelerator tables? How and why? The accelerator tables are > built from the debug info that's going into the final file, I think you're > conflating two problems here. > dtrace is munging through a bunch of code, some of it is the debug info. The accelerator tables are fubared going into the dtrace mode. Once I turned off the accel tables, the 'dwarfdump --verify' reported no errors. However, dtrace is still thrashing. I have to investigate via dtrace debugging and junk. But I still think this patch is worth keeping in for the time being, because it makes dwarfdump happy, which is certain to be an error during an Apple-style build... -bw _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
