On Wed, Dec 18, 2024 at 2:37 PM David Malcolm <dmalc...@redhat.com> wrote: > > On Tue, 2024-11-26 at 16:35 -0500, Marek Polacek wrote: > > On Mon, Nov 25, 2024 at 10:49:02PM -0500, David Malcolm wrote: > > > This patch attempts to provide better error messages for > > > code compiled with C23 that hasn't been updated for > > > "bool", "true", and "false" becoming keywords (based on > > > a brief review of the Gentoo bug tracker links given at > > > https://gcc.gnu.org/pipermail/gcc/2024-November/245185.html). > > > > > > Specifically: > > > > > > (1) with "typedef int bool;" previously we emitted: > > > > > > t1.c:7:13: error: two or more data types in declaration specifiers > > > 7 | typedef int bool; > > > | ^~~~ > > > t1.c:7:1: warning: useless type name in empty declaration > > > 7 | typedef int bool; > > > | ^~~~~~~ > > > > > > whereas with this patch we emit: > > > > > > t1.c:7:13: error: 'bool' cannot be defined via 'typedef' > > > 7 | typedef int bool; > > > | ^~~~ > > > t1.c:7:13: note: 'bool' is a keyword with '-std=c23' onwards > > > t1.c:7:1: warning: useless type name in empty declaration > > > 7 | typedef int bool; > > > | ^~~~~~~ > > > > > > (2) with "int bool;" previously we emitted: > > > > > > t2.c:7:5: error: two or more data types in declaration specifiers > > > 7 | int bool; > > > | ^~~~ > > > t2.c:7:1: warning: useless type name in empty declaration > > > 7 | int bool; > > > | ^~~ > > > > > > whereas with this patch we emit: > > > > > > t2.c:7:5: error: 'bool' cannot be used here > > > 7 | int bool; > > > | ^~~~ > > > t2.c:7:5: note: 'bool' is a keyword with '-std=c23' onwards > > > t2.c:7:1: warning: useless type name in empty declaration > > > 7 | int bool; > > > | ^~~ > > > > > > (3) with "typedef enum { false = 0, true = 1 } _Bool;" previously > > > we > > > emitted: > > > > > > t3.c:7:16: error: expected identifier before 'false' > > > 7 | typedef enum { false = 0, true = 1 } _Bool; > > > | ^~~~~ > > > t3.c:7:38: error: expected ';', identifier or '(' before '_Bool' > > > 7 | typedef enum { false = 0, true = 1 } _Bool; > > > | ^~~~~ > > > t3.c:7:38: warning: useless type name in empty declaration > > > > > > whereas with this patch we emit: > > > > > > t3.c:7:16: error: cannot use keyword 'false' as enumeration > > > constant > > > 7 | typedef enum { false = 0, true = 1 } _Bool; > > > | ^~~~~ > > > t3.c:7:16: note: 'false' is a keyword with '-std=c23' onwards > > > t3.c:7:38: error: expected ';', identifier or '(' before '_Bool' > > > 7 | typedef enum { false = 0, true = 1 } _Bool; > > > | ^~~~~ > > > t3.c:7:38: warning: useless type name in empty declaration > > > > > > > > > Successfully bootstrapped & regrtested on x86_64-pc-linux-gnu. > > > OK for trunk? > > > > Thanks for the patch. > > > > For: > > > > typedef int _Bool; > > > > or > > > > int _Bool; > > > > this patch says: > > > > note: '_Bool' is a keyword with '-std=c23' onwards > > > > but I don't think that's true: _Bool was added in C99 and is > > obsolescent in > > C23. > > Is _Bool meant to be usable in c90?
Yes. Since `_Bool` in the implementation reserved namespace (single underscore followed by a capital letter), it can be done even for C90. _Complex is in the same boat as _Bool here too but are in the implementation reserved namespaces. This is different from complex and bool, etc. We do have a -pedantic warning/error for its usage though. Thanks, Andrew > > I tried with -std=c90 and it fails; see: > https://godbolt.org/z/Wf6a9EKP9 > > <source>:1:13: error: two or more data types in declaration specifiers > 1 | typedef int _Bool; > | ^~~~~ > <source>:1:1: warning: useless type name in empty declaration > 1 | typedef int _Bool; > | ^~~~~~~ > > Thanks > Dave >
