[core] standardize on uint16_t for texture sizes

This commit is contained in:
Konstantin Käfer
2016-08-22 17:57:22 +02:00
parent 512037cfdc
commit 63e53de157
15 changed files with 42 additions and 36 deletions

View File

@@ -208,7 +208,7 @@ void NativeMapView::render() {
// take snapshot
const unsigned int w = fbWidth;
const unsigned int h = fbHeight;
mbgl::PremultipliedImage image { w, h };
mbgl::PremultipliedImage image { static_cast<uint16_t>(w), static_cast<uint16_t>(h) };
MBGL_CHECK_ERROR(glReadPixels(0, 0, w, h, GL_RGBA, GL_UNSIGNED_BYTE, image.data.get()));
const size_t stride = image.stride();
auto tmp = std::make_unique<uint8_t[]>(stride);