[Gnu-arch-users] more about GCC address@hidden: Re: Regressions on mainline]

[Tom Lord]

    From: "Joseph S. Myers" <address@hidden>
    To: Mark Mitchell <address@hidden>
    cc: address@hidden
    Subject: Re: Regressions on mainline

    > The log from http://gcc.gnu.org/regtest/ shows that on one platform with
    > an automated tester telling people about regressions it was in January
    > there were last no testsuite regressions.  Many regressions appear after a
    > period of build failure that makes identifying the problem patch more
    > difficult.

    > Only 251 out of 2260 lines in that log show no regressions, 789 show build
    > failures (this may be out of proportion to the time broken, as build
    > failures shorten the time for a test cycle) and 1215 show testsuite
    > regressions.  (Plus 5 other lines in the log file.)

Fascinating, isn't it?

It _sounds_ like (anyone privy to the details?) the initial integration
of a major patch went less than smoothly (SSA branch?).  Hence the
"Many regressions appear after a period of build failure...."

No conclusion stated here --- just trying to point out the kind of raw
data I try to pay attention to and, incidentally, wondering what
insights might be around to guide its interpretation.

-t