Hi Orgad, Orgad Shaneh <[email protected]> writes:
> On Fri, Sep 12, 2025 at 10:54 AM Paul Eggert <[email protected]> wrote: >> >> On 2025-09-11 22:34, Bruno Haible via Gnulib discussion list wrote: >> > What would be the point of this change? What does this fix or improve? >> >> I have the same question. >> >> More generally, many parts of Gnulib use a signed type (typically idx_t) >> for object sizes, so that arithmetic is less confusing and so that >> arithmetic overflow can be detected automatically. Using int, for sizes >> that fit into int, follows this practice. > > Hi Bruno and Paul, > > Thanks for your feedback. I understand the patch is not needed, and > I'll call it off. > > It was originally done by a colleague who couldn't recall the specific > reason for it, though it might have been an attempt to suppress a > compiler warning. > > I've checked with all our toolchains, and none showed such a warning. > > Sorry for the noise. > - Orgad For future reference there is compiler warnings listed in the HACKING of Gnulib's root directory for GCC 15 and Clang 20. These will disable some warnings that Gnulib does not care about, and some ones that are enabled by -Wall -Wextra. We typically silence false positives for those versions, but ignore them for older versions. Since it is a bit too difficult for us to track every compiler bug. :) Thanks, Collin
