Thread: CompilerWarnings task does not catch C++ warnings
The CompilerWarnings task on Cirrus CI does not catch warnings in C++ code. It tries to make warnings fatal by passing COPT='-Werror', but that does not apply to C++ compilations. I suggest that we just add COPT to CXXFLAGS as well. I think passing -Werror is just about the only reasonable use of COPT nowadays, so making that more robust seems useful. I don't think there is a need for a separate make variable for C++ here.
Attachment
Peter Eisentraut <peter@eisentraut.org> writes:
> The CompilerWarnings task on Cirrus CI does not catch warnings in C++
> code. It tries to make warnings fatal by passing COPT='-Werror', but
> that does not apply to C++ compilations.
> I suggest that we just add COPT to CXXFLAGS as well. I think passing
> -Werror is just about the only reasonable use of COPT nowadays, so
> making that more robust seems useful. I don't think there is a need for
> a separate make variable for C++ here.
+1, but what about the meson side of things?
regards, tom lane
On 18.06.24 16:08, Tom Lane wrote:
> Peter Eisentraut <peter@eisentraut.org> writes:
>> The CompilerWarnings task on Cirrus CI does not catch warnings in C++
>> code. It tries to make warnings fatal by passing COPT='-Werror', but
>> that does not apply to C++ compilations.
>> I suggest that we just add COPT to CXXFLAGS as well. I think passing
>> -Werror is just about the only reasonable use of COPT nowadays, so
>> making that more robust seems useful. I don't think there is a need for
>> a separate make variable for C++ here.
>
> +1, but what about the meson side of things?
If you use meson {setup|configure} --werror, that would affect both C
and C++ compilers.
Hi, On 2024-06-18 09:27:02 +0200, Peter Eisentraut wrote: > The CompilerWarnings task on Cirrus CI does not catch warnings in C++ code. > It tries to make warnings fatal by passing COPT='-Werror', but that does not > apply to C++ compilations. > > I suggest that we just add COPT to CXXFLAGS as well. I think passing > -Werror is just about the only reasonable use of COPT nowadays, so making > that more robust seems useful. I don't think there is a need for a separate > make variable for C++ here. +1 Greetings, Andres Freund