https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120990
Bug ID: 120990
Summary: Narrowing conversion in std::function return value is
not reported
Product: gcc
Version: 15.1.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: rohan.a.suri at gmail dot com
Target Milestone: ---
Created attachment 61815
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=61815&action=edit
Repro of hidden narrowing conversion with std::function
Hi team,
I'd like to report an undetected case of integer narrowing conversion with the
GCC compiler for C++.
A std::function can be defined to return an int32, but the actual lambda
implementation could be returning a larger integer say int64. In such a case,
there's a hidden narrowing conversion happening from int64 to int32 which
causes data loss.
I've attached a small repro of the issue to this report.
Here's also a live example to play around with on godbolt.org:
https://godbolt.org/z/G8xYh5c36
I have tried it on x86-64 (as well as ARM64) GCC 15.1 along with compiler
flags: -Wall -Wextra and no issues are reported.
Interestingly, if I try this on ARM 32-bit GCC 15.1, it does report:
<source>: In function 'size_t bar()':
<source>:7:12: warning: unsigned conversion from 'long long int' to 'size_t'
{aka 'unsigned int'} changes value from '10000000000' to '1410065408'
[-Woverflow]
7 | return 10000000000; // 10 billion
| ^~~~~~~~~~~
Please let me know if you need any more details from my side.
Thank you.
