[libcamera-devel] [PATCH 09/12] android: jpeg: Use maxJpegBufferSize() for compatibility

[Jacopo Mondi]

Platforms that do not provide a memory backend implementation should
keep using the maxJpegBufferSize() value to calculate the location where
to place the JPEG blob id, as the android_generic backend returns the
allocated buffer size as calculated using lseek which is larger than
the maximum JPEG frame size, which is where the framework expects the
JPEG blob to be placed.

Signed-off-by: Jacopo Mondi <jacopo at jmondi.org>
---
 src/android/jpeg/post_processor_jpeg.cpp | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/src/android/jpeg/post_processor_jpeg.cpp b/src/android/jpeg/post_processor_jpeg.cpp
index d6eeb962e81d..e7f66d66698c 100644
--- a/src/android/jpeg/post_processor_jpeg.cpp
+++ b/src/android/jpeg/post_processor_jpeg.cpp
@@ -185,9 +185,17 @@ int PostProcessorJpeg::process(const FrameBuffer &source,
 	}
 
 	/* Fill in the JPEG blob header. */
-	uint8_t *resultPtr = destination->plane(0) +
-			     destination->planeSize(0) -
-			     sizeof(struct camera3_jpeg_blob);
+	/*
+	 * \todo For backward compatibility reasons with the android_generic
+	 * memory backend, continue using the maxJpegBufferSize in case the
+	 * computed buffer size is larger. This can be dropped once all
+	 * supported platforms will have a working memory backend that
+	 * returns the correct buffer size.
+	 */
+	size_t blobSize = std::min<unsigned int>(cameraDevice_->maxJpegBufferSize(),
+						 destination->planeSize(0));
+	uint8_t *resultPtr = destination->plane(0) + blobSize
+			   - sizeof(struct camera3_jpeg_blob);
 	auto *blob = reinterpret_cast<struct camera3_jpeg_blob *>(resultPtr);
 	blob->jpeg_blob_id = CAMERA3_JPEG_BLOB_ID;
 	blob->jpeg_size = jpeg_size;
-- 
2.30.0