Adrian Prantl via llvm-dev
2021-Jul-15 23:47 UTC
[llvm-dev] [DebugInfo] Enabling constructor homing by default in clang
Thanks for letting us know. Is my understanding correct that "enabling by default" means when -flimit-debuginfo would be used, or would it also affect -fstandalone-debuginfo? As always, please watch the https://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake <https://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake> bot after making debug info changes. thanks, adrian> On Jul 15, 2021, at 11:14 AM, David Blaikie <dblaikie at gmail.com> wrote: > > Thanks Amy! + a few folks to highlight this. > > On Thu, Jul 15, 2021 at 11:06 AM Amy Huang via llvm-dev <llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>> wrote: > A while back I tried turning on constructor homing (-debug-info-kind=constructor) by default in clang. Ctor homing is a debug info optimization that reduces the amount of debug info that is emitted for class types. > > The main blocker was that for some libc++ types we avoid calling the constructor. Since then I've added a attribute for classes that ignores ctor homing (`__attribute__((standlone_debug))`) and added it to the libc++ types (7c2f58278e66). > > I think at this point we can try enabling it by default in Clang -- it's currently being used in chromium and google codebases. Thought I'd email the llvm-dev list in case people have thoughts/opinions/concerns about this. (patch is at https://reviews.llvm.org/D106084 <https://reviews.llvm.org/D106084>) > > _______________________________________________ > LLVM Developers mailing list > llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org> > https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev <https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev>-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20210715/2317c3fc/attachment.html>
Amy Huang via llvm-dev
2021-Jul-15 23:51 UTC
[llvm-dev] [DebugInfo] Enabling constructor homing by default in clang
Yep, it would happen with -flimit-debug-info (-fno-standalone-debug) but won't affect -fstandalone-debug. On Thu, Jul 15, 2021 at 4:47 PM Adrian Prantl <aprantl at apple.com> wrote:> Thanks for letting us know. Is my understanding correct that "enabling by > default" means when -flimit-debuginfo would be used, or would it also > affect -fstandalone-debuginfo? > > As always, please watch the > https://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake bot after > making debug info changes. > > thanks, > adrian > > On Jul 15, 2021, at 11:14 AM, David Blaikie <dblaikie at gmail.com> wrote: > > Thanks Amy! + a few folks to highlight this. > > On Thu, Jul 15, 2021 at 11:06 AM Amy Huang via llvm-dev < > llvm-dev at lists.llvm.org> wrote: > >> A while back I tried turning on constructor homing >> (-debug-info-kind=constructor) by default in clang. Ctor homing is a debug >> info optimization that reduces the amount of debug info that is emitted for >> class types. >> >> The main blocker was that for some libc++ types we avoid calling the >> constructor. Since then I've added a attribute for classes that ignores >> ctor homing (`__attribute__((standlone_debug))`) and added it to the libc++ >> types (7c2f58278e66). >> >> I think at this point we can try enabling it by default in Clang -- it's >> currently being used in chromium and google codebases. Thought I'd email >> the llvm-dev list in case people have thoughts/opinions/concerns about >> this. (patch is at https://reviews.llvm.org/D106084) >> _______________________________________________ >> LLVM Developers mailing list >> llvm-dev at lists.llvm.org >> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev >> > >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20210715/877fc933/attachment.html>