llvm dev - Feb 2016 - Possible soundness issue with available_externally (split from "RFC: Add guard intrinsics")

On Sat, Feb 27, 2016 at 1:41 PM, Sanjoy Das
<sanjoy at playingwithpointers.com> wrote:
> Just as a reality check, I wrote up a demonstration where one link
> order causes a SIGFPE and another doesn't (and the program is well
> defined, as far as I can tell).  All TUs are compiled with -O3.  This is
also
> an instance where we don't actually speculate an inline function, but
only
> DSE across it (after deducing readnone).
>
> Here's the link https://github.com/sanjoy/comdat-ipo
This test case "works" with gcc 5.3.0 too, afaict.

This is what I used:

Using built-in specs.
COLLECT_GCC=/usr/local/Cellar/gcc/5.3.0/bin/x86_64-apple-darwin15.0.0-g++-5
COLLECT_LTO_WRAPPER=/usr/local/Cellar/gcc/5.3.0/libexec/gcc/x86_64-apple-darwin15.0.0/5.3.0/lto-wrapper
Target: x86_64-apple-darwin15.0.0
Configured with: ../configure --build=x86_64-apple-darwin15.0.0
--prefix=/usr/local/Cellar/gcc/5.3.0
--libdir=/usr/local/Cellar/gcc/5.3.0/lib/gcc/5
--enable-languages=c,c++,objc,obj-c++,fortran --program-suffix=-5
--with-gmp=/usr/local/opt/gmp --with-mpfr=/usr/local/opt/mpfr
--with-mpc=/usr/local/opt/libmpc --with-isl=/usr/local/opt/isl
--with-system-zlib --enable-libstdcxx-time=yes
--enable-stage1-checking --enable-checking=release --enable-lto
--with-build-config=bootstrap-debug --disable-werror
--with-pkgversion='Homebrew gcc 5.3.0'
--with-bugurl=https://github.com/Homebrew/homebrew/issues
--enable-plugin --disable-nls --enable-multilib
Thread model: posix
gcc version 5.3.0 (Homebrew gcc 5.3.0)

-- Sanjoy

Nice repro.  It looks well-defined to me too.

Is there a PR open yet?
> On 2016-Feb-29, at 08:21, Sanjoy Das via llvm-dev <llvm-dev at
lists.llvm.org> wrote:
> 
> On Sat, Feb 27, 2016 at 1:41 PM, Sanjoy Das
> <sanjoy at playingwithpointers.com> wrote:
>> Just as a reality check, I wrote up a demonstration where one link
>> order causes a SIGFPE and another doesn't (and the program is well
>> defined, as far as I can tell).  All TUs are compiled with -O3.  This
is also
>> an instance where we don't actually speculate an inline function,
but only
>> DSE across it (after deducing readnone).
>> 
>> Here's the link https://github.com/sanjoy/comdat-ipo
> 
> This test case "works" with gcc 5.3.0 too, afaict.
> 
> This is what I used:
> 
> Using built-in specs.
> COLLECT_GCC=/usr/local/Cellar/gcc/5.3.0/bin/x86_64-apple-darwin15.0.0-g++-5
>
COLLECT_LTO_WRAPPER=/usr/local/Cellar/gcc/5.3.0/libexec/gcc/x86_64-apple-darwin15.0.0/5.3.0/lto-wrapper
> Target: x86_64-apple-darwin15.0.0
> Configured with: ../configure --build=x86_64-apple-darwin15.0.0
> --prefix=/usr/local/Cellar/gcc/5.3.0
> --libdir=/usr/local/Cellar/gcc/5.3.0/lib/gcc/5
> --enable-languages=c,c++,objc,obj-c++,fortran --program-suffix=-5
> --with-gmp=/usr/local/opt/gmp --with-mpfr=/usr/local/opt/mpfr
> --with-mpc=/usr/local/opt/libmpc --with-isl=/usr/local/opt/isl
> --with-system-zlib --enable-libstdcxx-time=yes
> --enable-stage1-checking --enable-checking=release --enable-lto
> --with-build-config=bootstrap-debug --disable-werror
> --with-pkgversion='Homebrew gcc 5.3.0'
> --with-bugurl=https://github.com/Homebrew/homebrew/issues
> --enable-plugin --disable-nls --enable-multilib
> Thread model: posix
> gcc version 5.3.0 (Homebrew gcc 5.3.0)
> 
> -- Sanjoy
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev

On Mon, Feb 29, 2016 at 9:22 AM, Duncan P. N. Exon Smith
<dexonsmith at apple.com> wrote:
> Nice repro.  It looks well-defined to me too.
>
> Is there a PR open yet?
Just filed PR26774.

-- Sanjoy

On Mon, Feb 29, 2016 at 8:21 AM, Sanjoy Das
<sanjoy at playingwithpointers.com> wrote:
> On Sat, Feb 27, 2016 at 1:41 PM, Sanjoy Das
> <sanjoy at playingwithpointers.com> wrote:
>> Just as a reality check, I wrote up a demonstration where one link
>> order causes a SIGFPE and another doesn't (and the program is well
>> defined, as far as I can tell).  All TUs are compiled with -O3.  This
is also
>> an instance where we don't actually speculate an inline function,
but only
>> DSE across it (after deducing readnone).
>>
>> Here's the link https://github.com/sanjoy/comdat-ipo
>
> This test case "works" with gcc 5.3.0 too, afaict.
I should have been clearer here -- by the above I meant gcc also
miscompiles the program in the same way.

-- Sanjoy