On Wed, Sep 27, 2017 at 6:58 PM, Richard Sandiford <richard.sandif...@linaro.org> wrote: > David Edelsohn <dje....@gmail.com> writes: >> On Fri, Sep 15, 2017 at 2:53 AM, Richard Biener <rguent...@suse.de> wrote: >>> On Thu, 14 Sep 2017, David Edelsohn wrote: >>> >>>> * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar >>>> to VN_TOP. >>>> >>>> This seems to have regressed >>>> >>>> FAIL: gcc.dg/tree-prof/time-profiler-2.c scan-ipa-dump-times profile >>>> "Read tp_first_run: 0" 2 >>>> FAIL: gcc.dg/tree-prof/time-profiler-2.c scan-ipa-dump-times profile >>>> "Read tp_first_run: 2" 1 >>>> FAIL: gcc.dg/tree-prof/time-profiler-2.c scan-ipa-dump-times profile >>>> "Read tp_first_run: 3" 1 >>> >>> Hmm, I don't see these FAILs. Looking at the testcase there are >>> no undefined uses so I wonder how the patch could have any effect. >>> >>> Can you re-check and open a bugreport? >> >> It disappeared again. A different failure appeared and disappeared a >> few weeks ago. Something in the testsuite infrastructure appears to >> not be stable, at least on AIX. Sorry for the incorrect report. > > Perhaps this is unrelated, but when doing the "has this patch > changed assembly on these targets?" testing, I noticed that AIX > had differences like: > > --- old/powerpc-ibm-aix7.0/test/-O3/g++.dg/init/constant1.s > +++ new/powerpc-ibm-aix7.0/test/-O3/g++.dg/init/constant1.s > @@ -4,21 +4,21 @@ > .csect ..text.startup[PR],2 > .align 2 > .align 4 > - .globl > _GLOBAL__I_65535_0_.._.._.._testsuite_g__.dg_._init_constant1.C_00000000_0xbb0d20e181e3a401 > - .globl > ._GLOBAL__I_65535_0_.._.._.._testsuite_g__.dg_._init_constant1.C_00000000_0xbb0d20e181e3a401 > - .csect > _GLOBAL__I_65535_0_.._.._.._testsuite_g__.dg_._init_constant1.C_00000000_0xbb0d20e181e3a401[DS] > -_GLOBAL__I_65535_0_.._.._.._testsuite_g__.dg_._init_constant1.C_00000000_0xbb0d20e181e3a401: > - .long > ._GLOBAL__I_65535_0_.._.._.._testsuite_g__.dg_._init_constant1.C_00000000_0xbb0d20e181e3a401, > TOC[tc0], 0 > + .globl > _GLOBAL__I_65535_0_.._.._.._testsuite_g__.dg_._init_constant1.C_00000000_0x5610f7ec143966c9 > + .globl > ._GLOBAL__I_65535_0_.._.._.._testsuite_g__.dg_._init_constant1.C_00000000_0x5610f7ec143966c9 > + .csect > _GLOBAL__I_65535_0_.._.._.._testsuite_g__.dg_._init_constant1.C_00000000_0x5610f7ec143966c9[DS] > +_GLOBAL__I_65535_0_.._.._.._testsuite_g__.dg_._init_constant1.C_00000000_0x5610f7ec143966c9: > + .long > ._GLOBAL__I_65535_0_.._.._.._testsuite_g__.dg_._init_constant1.C_00000000_0x5610f7ec143966c9, > TOC[tc0], 0 > > even though -frandom-seed is forced to the same value for both runs. > > Is this behaviour deliberate? I thought runs from a few weeks ago > had stable names, but maybe I just misremember.
AIX does not have init/fini sections, so it is built at link time as a C file by collect2-ld. I suspect that collect2-ld doesn't use -frandom-seed to build its temporary file. - David
