Lovell Fuller

Thanks, we'd need to expose this information via libvips first. The underlying TIFF "compression" value is an integer - see https://gitlab.com/libtiff/libtiff/-/blob/master/libtiff/tiff.h#L182-218 We'd probably want to map these to the relevant...

Please can you provide a sample image and code that allows someone else to reproduce.

Reducing the input dimensions by exactly 1/4 to reach the target dimensions, as in this example, is susceptible to the libjpeg-turbo shrink-on-load rounding error detailed at https://github.com/lovell/sharp/issues/3066. For this reason,...

We can't predict when libjpeg-turbo will differ in its rounding calculations (e.g. differing SIMD intrinsics) so must halve its use when a 1/2, 1/4, 1/8 shrink produces one of the...

> Are there examples where shrinks on an input image which would divide exactly (i.e. width and height are integer multiples of jpegShrinkOnLoad) could cause this issue? #3526 was one...

I cloned https://github.com/Aminelahlou/sharpCmykTest then ran `npm start` and it successfully generated the `output/test-sharp-output.tif` file. Your comments above mention an "Input buffer contains unsupported image format" error but I don't see...

Ah, thanks, I missed that bit. It looks like `whiteImage` contains raw pixel data. https://github.com/Aminelahlou/sharpCmykTest/blob/f524c80aa8e0e2666eeaabafd79ee9e299db9872/test-sharp-composite#L30-L40 Raw pixel data needs to be described when passed as input, e.g. with `width`, but...

You might need to add a specific CMYK ICC profile to intermediary images via [withMetadata](https://sharp.pixelplumbing.com/api-output#withmetadata) to avoid the built-in CMYK profile being used for any implicit RGB to CMYK conversions.

libvips provides a `compositing_space` option when compositing that sharp currently leaves with the default setting. https://www.libvips.org/API/current/libvips-conversion.html#vips-composite > This is VIPS_INTERPRETATION_sRGB, VIPS_INTERPRETATION_B_W, VIPS_INTERPRETATION_RGB16, or VIPS_INTERPRETATION_GREY16 by default As a future possible...

The `affine()` operation is for applying matrix transformations and accepts floats. Did you see the `idx` etc. properties for translation offsets? https://sharp.pixelplumbing.com/api-operation#affine