Re: CI with clang 18 ?

2024-07-13 Thread Bruno Haible
I wrote: > so I can know why the CI did not report it within three weeks That was because the compilation error occurred only with clang 18, which was released in March 2024 and is therefore not yet present in the *BSD distros. In the last two years, I got the feeling that clang is adopting new I

Re: CI with clang 18 ?

2024-07-13 Thread Paul Eggert
On 7/13/24 09:15, Bruno Haible wrote: In the last two years, I got the feeling that clang is adopting new ISO C features more quickly than GCC (seen with the K&R prototypes removal in clang 16, now with ckd_add in clang 18). I don't have that feeling. I would put it differently, since GCC has

tests: Don't depend on the exact locale name

2024-07-13 Thread Bruno Haible
The locale names that I hardcoded in some tests are not those present in some systems. This patch fixes it. 2024-07-13 Bruno Haible tests: Don't depend on the exact locale name. * tests/test-c-ctype.c: Include . (main): Test the return value of setlocale(). Try not onl

doc: Revamp list of dependencies

2024-07-13 Thread Bruno Haible
This patch improves the DEPENDENCIES file, mainly by distinguishing the maintainer environment and the target environment. 2024-07-13 Bruno Haible doc: Revamp list of dependencies. * DEPENDENCIES: Clarify that GNU make is only needed on OpenBSD and Solaris. Mark GNU ta

Re: CI with clang 18 ?

2024-07-13 Thread Collin Funk
Bruno Haible writes: > In contrast, a CI build with the newest released gcc version seems > less urgent, since > - Fedora jumps on the newest gcc quickly, > - gcc is not so frequently ahead of clang regarding ISO C language > features. > Right? > > Opinions? Not sure about language stand

Re: CI with clang 18 ?

2024-07-13 Thread Jim Meyering
On Sat, Jul 13, 2024 at 12:16 AM Bruno Haible wrote: > I wrote: > > so I can know why the CI did not report it within three weeks > > That was because the compilation error occurred only with clang 18, > which was released in March 2024 and is therefore not yet present > in the *BSD distros. > > I

Re: CI with clang 18 ?

2024-07-13 Thread Bruno Haible
Jim Meyering wrote: > It feels like a good thing to do, but will come with some cost: I have > the impression that relatively few GNU projects use clang (and esp > clang 18), so this testing is likely to expose more "issues". However, > a more diverse set of compilers/warnings may expose real probl

Re: CI with clang 18 ?

2024-07-13 Thread Bruno Haible
Collin Funk wrote: > Not sure about language standards but I know that Clang releases > sometimes introduce some warnings/errors which might be nice to see in > the logs. > > It has been a while but I remember finding -Wtautological-* and > -Wbitwise-instead-of-logical to be annoying when they wer

Re: CI with clang 18 ?

2024-07-13 Thread Bruno Haible
Paul Eggert wrote: > > In the last two years, I got the feeling that clang is adopting new > > ISO C features more quickly than GCC (seen with the K&R prototypes > > removal in clang 16, now with ckd_add in clang 18). > > I don't have that feeling. I would put it differently, since GCC has > supp

Re: CI with clang 18 ?

2024-07-13 Thread Collin Funk
Bruno Haible writes: > There are clang+llvm binaries for 18.1.x [1] for x86_64-linux-gnu-ubuntu, > which is perfect for the CI machines (that are currently equipped with > Ubuntu 22.04). For other glibc platforms, I would build clang+llvm from > source. For *BSD platforms, I would wait on their "

stdlib: Support use of clang++ on Ubuntu

2024-07-13 Thread Bruno Haible
There are two ways to use the pre-built binaries of clang+llvm on a glibc system, such as Ubuntu: a) Use CC=clang, CXX=clang++. This makes use of the clang++ runtime libraries. But AFAIU these are ABI-incompatible with the system's libstdc++ and thus with all other preinstalled librari

Re: compiling LLVM

2024-07-13 Thread Bruno Haible
Collin Funk wrote: > Though my machine always ran out of memory when compiling > LLVM unless I did 'ninja -j 1' so be careful. My system might just be > showing its age though. It's not your system's age that is the problem. It's the C++ toolchains (compiler and linker) that take a huge amount of

announce-gen: Fix copyright year in --version output.

2024-07-13 Thread Collin Funk
The announce-gen and useless-if-before-free scripts have had changes since their copyright date was updated: $ ./build-aux/announce-gen --version announce-gen version 2024-07-04 10:55 Copyright (C) 2022 Free Software Foundation, Inc. I've pushed the two attached patches to derive the

Re: announce-gen: Fix copyright year in --version output.

2024-07-13 Thread Bruno Haible
Hi Collin, > I've pushed the two attached patches to derive the copyright year from > the time-stamp string. > +(my $copyright_year = $VERSION) =~ s/^(\d*)-.*$/$1/; Can this be rewritten in a simpler way, that does not use idioms from perl's bag of tricks? Something like my $copyright_year =

Re: announce-gen: Fix copyright year in --version output.

2024-07-13 Thread Collin Funk
Hi Bruno, Bruno Haible writes: > Can this be rewritten in a simpler way, that does not use idioms > from perl's bag of tricks? Something like > > my $copyright_year = ; > > like one would do it in every other programming language? Not sure. I'm not the most knowledgeable perl programmer. Mayb