On Mon, Jan 21, 2013 at 9:41 AM, Eric Botcazou <ebotca...@adacore.com> wrote: > Hi, > > this is a regression present in the Ada compiler since 4.5: the issue had been > latent for ages, but an unrelated streamlining of the IR made it appear. > > When make_range_step is invoked on: > > (integer)!b < 0 > > where b is a boolean, it returns "always true" instead of "always false". > > The sequence is as follows: > > (integer)!b < 0 is_true_if not in [0:0] > (integer)!b is_true_if not in [0;+inf[ > !b is_true_if not in [0;255] > b is_true_if in [0;255] > > The wrong step is the last one: when TRUTH_NOT_EXPR is seen in make_range_step > the "in" value is unconditionally toggled. Of course that doesn't work in the > general case, just if the range is the "boolean" range. As a matter of fact, > this is explained just below for the comparison operators: > > /* We can only do something if the range is testing for zero > and if the second operand is an integer constant. Note that > saying something is "in" the range we make is done by > complementing IN_P since it will set in the initial case of > being not equal to zero; "out" is leaving it alone. */ > > so the fix is to use the zero range condition in the TRUTH_NOT_EXPR case. > > Tested on x86_64-suse-linux, OK for mainline? And for branch(es)?
Ok everywhere. Thanks, Richard. > > 2013-01-21 Eric Botcazou <ebotca...@adacore.com> > > * fold-const.c (make_range_step) <TRUTH_NOT_EXPR>: Bail out if the > range isn't testing for zero. > > > 2013-01-21 Eric Botcazou <ebotca...@adacore.com> > > * gnat.dg/opt26.adb: New test. > > > -- > Eric Botcazou
