Bug#780718: return in ?:-operator with function calls is incorrectly optimized
On 03/18/2015 11:39 AM, David Kalnischkies wrote:
> Package: g++
> Version: 4:4.7.2-1
> Severity: normal
>
> Hi,
>
> while testing apt with gcc-5 suite I noticed that while it compiles just
> fine, the resulting binaries produce incorrect results failing our
> integration tests (which aren't run at package buildtime, so this wasn't
> detected in the gcc-5 mass-rebuild).
>
> While trying to come up with a smaller testcase I noticed¹ that with
> a slight variation even the g++-4.x series² emits the same strange
> results… clang on the other hand produces the expected results, neither
> of them generates any warnings about the code.
>
>
> Attached is as far as I got in terms of a testcase. Still depends on
> libapt as if I make the method directly available everything is fine…
would it be possible to have a test case which doesn't link the library, maybe
by including all relevant files so that we have a single translation unit?
Reply to: