#4174 Pointer comparisons bypass compatible-type and object-type constraints

[Christopher Bazley]

1: Sample code that reproduces the problem.

int f(int *p, char *q) { return p == q; }

2: Exact command used to run SDCC on this sample code.

sdcc -mz80 --std=c2y --nostdinc -S bug.c -o bug.asm

3: SDCC version tested.

SDCC 4.6.4 #17064 (Linux), tested 21 October 2026. No candidate patch was applied.

4: Observed and expected behaviour.

SDCC exits successfully and emits no diagnostic for the sample.

WG14 N3886,
6.5.9 paragraph 2 and 6.5.10 paragraph 2, specifies the constraints violated by these examples.

int and char are incompatible types, so comparisons of these pointer types
require a diagnostic for ==, != and <. The special equality permission
involving void * does not apply to int * versus char *. Relational operators
require pointers to compatible object types; two function pointers therefore
cannot be ordered with <. Compatible object-pointer ordering, equality after
adding const, function-pointer equality and object-pointer/void-pointer
equality are valid cases. The valid ordering example checks translation
constraints; it makes no claim that arbitrary runtime arguments permit a
defined ordering.

5: Additional checks.

The sample is silently accepted on Z80, MCS51 default, STM8 default, PDK14 and MOS65C02 in C2Y mode. These are translation and diagnostic checks; no runtime result is claimed.

Compiler Explorer: SDCC 4.6.0, SDCC trunk and GCC 16.2. All three panes use the sample above. The saved source, compiler selections and options were round-trip verified on 21 October 2026. Both SDCC panes use -mz80 --std=c2y; GCC uses -std=c2y -pedantic-errors -fsyntax-only. The public trunk build is r17013, distinct from the local r17064 retest.

CE compiler	Observed result
SDCC 4.6.0	Accepted silently
SDCC trunk r17013	Accepted silently
GCC 16.2	Rejected with diagnostics

Case	Expected	Current SDCC results
compatible-order-control.c	Valid	Accepted silently on all five targets
function-pointer-control.c	Valid	Accepted silently on all five targets
function-pointer-order.c	Diagnostic required	Accepted silently on all five targets
pointer-equal.c	Diagnostic required	Accepted silently on all five targets
pointer-order.c	Diagnostic required	Accepted silently on all five targets
pointer-unequal.c	Diagnostic required	Accepted silently on all five targets
qualified-pointer-control.c	Valid	Accepted silently on all five targets
void-pointer-control.c	Valid	Accepted silently on all five targets

Related report: [bugs:#3836]. That report concerns excessive diagnostics for address-space comparisons; this report concerns missing diagnostics for incompatible ordinary pointed-to types.

[Philipp Klaus Krause]

IMO, this should be a warning.

[Christopher Bazley]

IMO, this should be a warning.

Yes. If it has never provoked an error before then it seems likely that existing programs violate the constraint. I implemented the diagnostic message as a warning in the attached patch. Please take a look when you have time, @spth.

[Philipp Klaus Krause]

The patch looks good, but I didn't check if the new warnings trigger in any existing regression tests; if they do we probably want #pragma disable_warning in them.

[Christopher Bazley]

The patch looks good, but I didn't check if the new warnings trigger in any existing regression tests; if they do we probably want #pragma disable_warning in them.

Yes, I needed to add that pragma in support/regression/tests/bug-2710.c. Thanks for the review and reminder.

[Christopher Bazley]

Fixed by [r17079]