[PATCH] libcamera: matrix: Fix compilation error in inverse() function

[Kieran Bingham]

Quoting Laurent Pinchart (2025-05-22 13:42:11)
> Some gcc versions report uninitialized variable usage:
> 
> In member function ‘constexpr T& libcamera::Span<T, 4294967295>::operator[](size_type) const [with T = unsigned int]’,
>     inlined from ‘void libcamera::matrixInvert(Span<const T>, Span<T, 4294967295>, unsigned int, Span<T, 4294967295>, Span<unsigned int>)::MatrixAccessor::swap(unsigned int, unsigned int) [with T = float]’ at ../../src/libcamera/matrix.cpp:194:13,
>     inlined from ‘bool libcamera::matrixInvert(Span<const T>, Span<T, 4294967295>, unsigned int, Span<T, 4294967295>, Span<unsigned int>) [with T = float]’ at ../../src/libcamera/matrix.cpp:255:14:
> ../../include/libcamera/base/span.h:362:76: error: ‘row’ may be used uninitialized [-Werror=maybe-uninitialized]
>   362 |         constexpr reference operator[](size_type idx) const { return data()[idx]; }
>       |                                                                      ~~~~~~^
> ../../src/libcamera/matrix.cpp: In function ‘bool libcamera::matrixInvert(Span<const T>, Span<T, 4294967295>, unsigned int, Span<T, 4294967295>, Span<unsigned int>) [with T = float]’:
> ../../src/libcamera/matrix.cpp:232:30: note: ‘row’ was declared here
>   232 |                 unsigned int row;
>       |                              ^~~
> 
> This is a false positive. Fix it by initializing the variable when
> declaring it.
> 
> Fixes: 6287ceff5aba ("libcamera: matrix: Add inverse() function")
> Signed-off-by: Laurent Pinchart <laurent.pinchart at ideasonboard.com>
> ---
>  src/libcamera/matrix.cpp | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/src/libcamera/matrix.cpp b/src/libcamera/matrix.cpp
> index ed22263b58f8..b7c07e896538 100644
> --- a/src/libcamera/matrix.cpp
> +++ b/src/libcamera/matrix.cpp
> @@ -229,7 +229,7 @@ bool matrixInvert(Span<const T> dataIn, Span<T> dataOut, unsigned int dim,
>                  * Locate the next pivot. To improve numerical stability, use
>                  * the row with the largest value in the pivot's column.
>                  */
> -               unsigned int row;
> +               unsigned int row = pivot;


Seems to make sense, and workaround the toolchain.

Reviewed-by: Kieran Bingham <kieran.bingham at ideasonboard.com>

>                 T maxValue{ 0 };
>  
>                 for (unsigned int i = pivot; i < dim; ++i) {
> 
> base-commit: efdbe3969841e342c30dfdced38b6ad9ad55dccf
> -- 
> Regards,
> 
> Laurent Pinchart
>