Jonathan Wakely wrote:
On Fri, 27 Nov 2020 at 15:57, emre brookes <brookes@xxxxxxxxxxx> wrote:
emre brookes wrote:
Martin Sebor wrote:
On 11/22/20 8:34 AM, emre brookes wrote:
I was compiling a project and received an odd message that took some
digging.
e.g.
file.cpp:#:#: error: expected unqualified-id before numeric constant
# | if ( inttype == MyClass::ERROR )
^~~~~
where MyClass::ERROR is an enum.
Took some digging to find out ERROR was somewhere #define'd and the
cure was to #undef ERROR
Could never find out *where* it was defined - recursively searched
all include files of the project and /usr/include etc.
Regardless, it would have been very helpful to know in the error
messages that a #define was active or to have the ERROR replaced by
the macro substitution.
Is it possible to get any macro expansion information in error
messages?
I didn't see a compiler flag for to get this information, but I
might have missed it.
If not, it would be a nice feature to have.
I'd expect to see output similar to the below where the error points
to the macro. I'm not sure under what conditions it won't (with some
specific options perhaps?) If you can create an isolated test case
showing it doesn't I'd recommend to open a bug in Bugzilla (or just
follow up here).
Martin
$ cat a.C && gcc -S a.C
struct MyClass { enum { ERROR = 1 }; };
#define ERROR 1
int f (int inttype)
{
if (inttype == MyClass::ERROR)
return 0;
return 1;
}
a.C: In function ‘int f(int)’:
a.C:3:15: error: expected unqualified-id before numeric constant
3 | #define ERROR 1
| ^
a.C:7:27: note: in expansion of macro ‘ERROR’
7 | if (inttype == MyClass::ERROR)
| ^~~~~
a.C:7:25: error: expected ‘)’ before numeric constant
7 | if (inttype == MyClass::ERROR)
| ~ ^
| )
.
I get the same results for your example code. Here gcc nicely
identifies the macro expansion :)
Will test later with compiler flags from the project's Makefile add
see if I can reproduce the issue I observed on a simple case.
Thanks,
Emre
Followup:
strange error
555 0 mingw [eb@ebwin764] ~/gccerrortest $ gcc --version
gcc.exe (Rev5, Built by MSYS2 project) 10.2.0
key takeway -
on msys2 64 bit
macro expansion information is lost when the #define ERROR is in an
include file in /mingw64/x86_64-w64-mingw32/include
if i move the include file elsewhere (tested /usr/include and .) -
normal behavior
linux variants tested did not show this behavior
centos 8 gcc (GCC) 9.2.1 20191120 (Red Hat 9.2.1-2)
centos 8 gcc (GCC) 8.3.1 20191121 (Red Hat 8.3.1-5)
ubuntu 20 gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0
Should I create a bugzilla for gcc? Might be some msys2 issue. Not sure.
I think this is the expected behaviour when the macro is defined in a
system header, although I'm not convinced it's useful. I think there's
already a bugzilla about it.
Ah, thank you! system headers, as in
https://gcc.gnu.org/onlinedocs/gcc-10.2.0/cpp/System-Headers.html
This explains the behavior I am seeing and can duplicate the issue using
-isystem as evidenced below on all tested versions of gcc.
The world makes sense again :)
It would be useful if involved system header macro expansion were noted
in compile error output.
Evidence:
$ cat a.h
#define ERROR 0
$ cat a.C
struct MyClass { enum { ERROR = 1 }; };
#include <a.h>
int f (int inttype)
{
if (inttype == MyClass::ERROR)
return 0;
return 1;
}
$ gcc -isystem . a.C
In file included from a.C:3:
a.C: In function ‘int f(int)’:
a.C:7:28: error: expected unqualified-id before numeric constant
if (inttype == MyClass::ERROR)
^~~~~
a.C:7:26: error: expected ‘)’ before numeric constant
if (inttype == MyClass::ERROR)
~ ^
)
