Hi Paul, > Le 22 oct. 2019 à 20:10, Paul Eggert <egg...@cs.ucla.edu> a écrit : > > On 10/21/19 2:07 AM, Akim Demaille wrote: > >>> I'd rather not. Newer compilers are doing their jobs better in this area, >> I don't see that, given that I have the problem with GCC 7, 8 and 9 on macOS. > > That's odd. What problems are these?
Everything I already reported :) See https://lists.gnu.org/archive/html/bug-bison/2019-10/msg00061.html and https://lists.gnu.org/archive/html/bug-bison/2019-10/msg00063.html both being about GCC9 > GCC 9 should have the necessary primitives to bypass the code eliciting the > warnings you mentioned; besides, recent GCCs are better at avoiding false > alarms with options like -Wtype-limits. I don't see why macOS GCC 9 would > differ from GNU/Linux GCC 9 in this area. I know. That's why I wrote >> Except that on this mac GCC 7 to 9 shows the warning. I don't understand >> [why] there's a difference with GNU/Linux. Maybe we should compare our outputs with -E of the small example I used in the second URL? > I just built Bison master with the attached patch (./configure > --enable-gcc-warnings --disable-nls), and "make" and "make check" both > succeeded on Fedora 30 with GCC 9.2.1 20190827 (Red Hat 9.2.1-1). How does > such a build fail for you with macOS 9 and GCC 9.2? > <type-limits.diff> Yes, this is 9.2.0 and I have back all the warnings reported in the first URL.