https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101862
Bug ID: 101862
Summary: [C, C++] Potential '?:' diagnostic for always-true
expressions in boolean context
Product: gcc
Version: 12.0
Status: UNCONFIRMED
Keywords: diagnostic
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: tschwinge at gcc dot gnu.org
Target Milestone: ---
Created attachment 51287
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51287&action=edit
'c-c++-common/goacc/prN.c'
Would it be possible for GCC to diagnose the "'?:' misuse" in the attached
C/C++ test case? No matter what gets stored in 'arr', the 'assert' never
triggers, because what's actually meant here, is:
for (int i = 0; i < 32; i++)
- assert (arr[i] == ((i % 2) != 0) ? i + 1 : i + 2);
+ assert (arr[i] == (((i % 2) != 0) ? i + 1 : i + 2));
This is going to be more complicated than
'gcc/c-family/c-common.c:c_common_truthvalue_conversion', 'case COND_EXPR:',
which diagnoses "'?:' using integer constants in boolean context" for
'INTEGER_CST's -- which these are not, of course. I suppose we'd need some
value range analysis (so, at some later stage in the pass pipeline?) to see
that 'i + 1'/'i + 2' are always-true expressions in boolean context?
Is this feasible or not? (..., and if yes, if anybody got any pointers about
where/how to do this...)
