On 09/02/2016 12:53 PM, Bernd Edlinger wrote:
Hi!
As reported in PR77434 and PR77421 there should be a warning for
suspicious uses of conditional expressions with non-boolean arguments.
This warning triggers on conditional expressions in boolean context,
when both possible results are non-zero integer constants, so that
the resulting truth value does in fact not depend on the condition
itself. Thus something like "if (a == b ? 1 : 2)" is always bogus,
and was most likely meant to be "if (a == (b ? 1 : 2))".
Boot-strap and reg-testing on x86_64-pc-linux-gnu without regressions.
Is it OK for trunk.
Thanks
Bernd.
changelog-pr77434.txt
gcc:
2016-09-02 Bernd Edlinger <bernd.edlin...@hotmail.de>
PR c++/77434
* doc/invoke.texi: Document -Wcond-in-bool-context.
PR middle-end/77421
* dwarf2out.c (output_loc_operands): Fix assertion.
c-family:
2016-09-02 Bernd Edlinger <bernd.edlin...@hotmail.de>
PR c++/77434
* c.opt (Wcond-in-bool-context): New warning.
* c-common.c (c_common_truthvalue_conversion): Warn on integer
constants in boolean context.
testsuite:
2016-09-02 Bernd Edlinger <bernd.edlin...@hotmail.de>
PR c++/77434
* c-c++-common/Wcond-in-bool-context.c: New test.
For some reason the non-symmerty of the changes to
c_common_truthvalue_conversion caused me to have to think far more about
this than I probably should have.
Couldn't we have a new function
integer_zerop_or_onep
Then use
&& (!integer_zerop_or_onep (TREE_OPERAND (expr, 1))
|| !integer_zerop_or_onep (TREE_OPERAND (expr, 2)))
Ie, if they're both constants and either is not [0,1], then we warn.
With that cleanup, this is OK.
jeff
