Joerg Sonnenberger via llvm-dev
2015-Sep-28 00:49 UTC
[llvm-dev] Error compiling libc++ for ARMv6
On Sun, Sep 27, 2015 at 05:07:17PM -0700, JF Bastien wrote:> It's part of the stable Linux ABI for ARM. Yes there are non-Linux OSes, > and I wasn't suggesting one emit these calls for non-Linux. > > Please clarify the red warning lights. FWIW these functions were used by > Chrome on ARM until very recently, specifically for the sake of older ARM > CPUs.I know. It is painful to find such code, especially since most instances never checked for Linux in first place. I'm waiting for a case where using libgcc calls doesn't work. Joerg
I'm still struggling to understand your specific objection. It sounds like you have something in mind, as well as a proposed alternative, could you describe that? Checking for Linux in the triple is indeed a requirement! I'd like to discuss concrete objections beyond being Linux specific. On Sep 27, 2015 5:49 PM, "Joerg Sonnenberger" <joerg at britannica.bec.de> wrote:> On Sun, Sep 27, 2015 at 05:07:17PM -0700, JF Bastien wrote: > > It's part of the stable Linux ABI for ARM. Yes there are non-Linux OSes, > > and I wasn't suggesting one emit these calls for non-Linux. > > > > Please clarify the red warning lights. FWIW these functions were used by > > Chrome on ARM until very recently, specifically for the sake of older ARM > > CPUs. > > I know. It is painful to find such code, especially since most instances > never checked for Linux in first place. I'm waiting for a case where > using libgcc calls doesn't work. > > Joerg >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150927/59a871e2/attachment.html>
Joerg Sonnenberger via llvm-dev
2015-Sep-28 11:44 UTC
[llvm-dev] Error compiling libc++ for ARMv6
On Sun, Sep 27, 2015 at 06:39:40PM -0700, JF Bastien wrote:> I'm still struggling to understand your specific objection. It sounds like > you have something in mind, as well as a proposed alternative, could you > describe that? > > Checking for Linux in the triple is indeed a requirement! I'd like to > discuss concrete objections beyond being Linux specific.Do not add anything in Clang *itself* that calls the kernel functions directly. Do not add any calls without Linux conditionals to any runtime library in LLVM. That's pretty much the whole point I want to make. Joerg