[Bug tree-optimization/88011] [9 regression] r266028 causes a bunch of go failures

[ian at airs dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88011

Ian Lance Taylor <ian at airs dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #45013|0                           |1
        is obsolete|                            |
  Attachment #45018|0                           |1
        is obsolete|                            |

--- Comment #8 from Ian Lance Taylor <ian at airs dot com> ---
Created attachment 45019
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=45019&action=edit
Standalone test case

An even smaller test case.  The miscompilation is definitely in the lehmerGCD
function.  The computation `ua+q*ub` is being dropped in the 115t.dom2 pass.  I
don't know why.  I haven't been able to recreate this in a C test case.  I
don't know why.

Note that in Go the division a / b compiles to
    a == -1 ? (b == -1 ? 1 : 0) : a / b
and a % b compiles to
    a == -1 ? (b == -1 ? 0 : a) : a % b
That is related to this, in that it causes extra basic blocks, and something
about the block merging in dom2 is causing the expression to be discarded.