Eli Friedman <eli.friedman at gmail.com> writes:> On Fri, Jul 8, 2011 at 9:13 AM, David Greene <dag at cray.com> wrote: >> I'm seeing a bunch of LLVM test failures that look like this: >> >> Exit Code: 1 >> Command Output (stderr): >> -- >> LLVM ERROR: SSE register return with SSE disabled >> >> Anyone else seeing this? > > That message means you're on 64-bit with SSE2 turned off. Probably > fallout from recent target-feature work, but none of the builders show > any issues; can you give more info?What would help? I simply updated, configure --prefix=<foo>, make, make check. $ uname -a Linux lamba 2.6.37.3-0.3-default #1 SMP Thu Apr 28 09:24:02 CDT 2011 x86_64 x86_64 x86_64 GNU/Linux -Dave
On Fri, Jul 8, 2011 at 11:02 AM, David A. Greene <greened at obbligato.org> wrote:> Eli Friedman <eli.friedman at gmail.com> writes: > >> On Fri, Jul 8, 2011 at 9:13 AM, David Greene <dag at cray.com> wrote: >>> I'm seeing a bunch of LLVM test failures that look like this: >>> >>> Exit Code: 1 >>> Command Output (stderr): >>> -- >>> LLVM ERROR: SSE register return with SSE disabled >>> >>> Anyone else seeing this? >> >> That message means you're on 64-bit with SSE2 turned off. Probably >> fallout from recent target-feature work, but none of the builders show >> any issues; can you give more info? > > What would help? I simply updated, configure --prefix=<foo>, make, make > check. > > $ uname -a > Linux lamba 2.6.37.3-0.3-default #1 SMP Thu Apr 28 09:24:02 CDT 2011 > x86_64 x86_64 x86_64 GNU/LinuxJust managed to reproduce this myself; looking... -Eli
Eli Friedman <eli.friedman at gmail.com> writes:> On Fri, Jul 8, 2011 at 11:02 AM, David A. Greene <greened at obbligato.org> wrote: >> Eli Friedman <eli.friedman at gmail.com> writes: >> >>> On Fri, Jul 8, 2011 at 9:13 AM, David Greene <dag at cray.com> wrote: >>>> I'm seeing a bunch of LLVM test failures that look like this: >>>> >>>> Exit Code: 1 >>>> Command Output (stderr): >>>> -- >>>> LLVM ERROR: SSE register return with SSE disabled > > Just managed to reproduce this myself; looking...I believe Evan's changes to Subtarget information is the culprit. I think you noticed that as well. -Dave