David Blaikie
2014-Jul-10 22:51 UTC
[LLVMdev] Quick question about your "Recommit r212203"
On Jul 10, 2014 11:44 AM, "Keno Fischer" <kfischer at college.harvard.edu> wrote:> > Yes, it's our own frontend (julia). I am running with assertions enabled,but I'm not getting any. It is perfectly possible that we're doing something incorrect. I wouldn't be surprised. Anything I can do to check? Hmm - are you putting your subprograms in the subprogram list of the compile unit in the metadata? Do you use DIBuilder to build your debug info?> Also, would it make sense to cc this to the mailing list now?Sure, added llvm-dev to the thread now.> > > On Thu, Jul 10, 2014 at 11:36 AM, David Blaikie <dblaikie at gmail.com>wrote:>> >> On Thu, Jul 10, 2014 at 11:12 AM, Keno Fischer >> <kfischer at college.harvard.edu> wrote: >> > Hi David, >> > >> > after your "Recommit r212203" commit, I suddenly don't have debug infoin my>> > object file any more. I don't quite understand the code well enough toknow>> > why. Any ideas? >> >> (you can email these questions to the mailing list - that enssures >> other people can benefit from the answers, etc) >> >> I'm going to guess/assume that you're generating IR from your own >> frontend? Perhaps you're generating something incorrect somewhere, but >> it's hard to guess. the recommit of r212203 should've just caused >> assertions (are you building/running LLVM with assertions enabled?) >> rather than any change in behavior, as such. > >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140710/82fc0939/attachment.html>
Hi David, I figured it out fairly quickly (I missed an argument to createFunction). It ended up taking a while longer to actually verify that it worked, because it exposed another bug on my dev machine where MCJIT would drop _eh_frame on mac, but that's unrelated. Thanks for getting back to me so quickly on this earlier! Keno On Thu, Jul 10, 2014 at 3:51 PM, David Blaikie <dblaikie at gmail.com> wrote:> > On Jul 10, 2014 11:44 AM, "Keno Fischer" <kfischer at college.harvard.edu> > wrote: > > > > Yes, it's our own frontend (julia). I am running with assertions > enabled, but I'm not getting any. It is perfectly possible that we're doing > something incorrect. I wouldn't be surprised. Anything I can do to check? > > Hmm - are you putting your subprograms in the subprogram list of the > compile unit in the metadata? Do you use DIBuilder to build your debug info? > > > Also, would it make sense to cc this to the mailing list now? > > Sure, added llvm-dev to the thread now. > > > > > > > On Thu, Jul 10, 2014 at 11:36 AM, David Blaikie <dblaikie at gmail.com> > wrote: > >> > >> On Thu, Jul 10, 2014 at 11:12 AM, Keno Fischer > >> <kfischer at college.harvard.edu> wrote: > >> > Hi David, > >> > > >> > after your "Recommit r212203" commit, I suddenly don't have debug > info in my > >> > object file any more. I don't quite understand the code well enough > to know > >> > why. Any ideas? > >> > >> (you can email these questions to the mailing list - that enssures > >> other people can benefit from the answers, etc) > >> > >> I'm going to guess/assume that you're generating IR from your own > >> frontend? Perhaps you're generating something incorrect somewhere, but > >> it's hard to guess. the recommit of r212203 should've just caused > >> assertions (are you building/running LLVM with assertions enabled?) > >> rather than any change in behavior, as such. > > > > >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140712/d7f58931/attachment.html>