Resizer bug in AVISynth 2.5.1 RC4

[RB]

Hi,

with latest RC4 Final of AVISynth 2.5.1, the following script

BilinearResize(688,432,17,73,688,426)
AddBorders(16,72,16,72)

results in an "Crop: YUV images can only be cropped by even numbers (left side)" error message. However, the Resizers should allow for fractional cropping according to the docs. IIRC this was working fine in previous releases of AVS 2.5.1.

[sh0dan]

When you don't resize (in one or both directions), cropping is passed on to the crop function, without using the resizer - this results in the best image quality, and fastest operation.

It has been so forever - but it could be changed...?!?

[Richard Berg]

I agree - fractional crops should only be done when AVS is forced to resample already by a resize operation. I'll bet your video had different dimensions before. (If not, please find the exact version in which there was different behavior.)

[Wilbert]

I'll bet your video had different dimensions before.
Are you sure, you want to bet :) ?

edit: :stupid: (guess it was to early)

[sh0dan]

Just to show you why:

>BilinearResize(688,432,17,73,688,426)

You crop to width 688, and resize to a width of 688 afterwards - so there is no need to resize - therefore resize passes the args on to crop.

[RB]

Yeah, I missed that horizontal resolution remains the same in my example, sorry. But still, shouldn't the docs be updated then to say that the resizers allow fractional cropping only if the resizer actually needs to resize?

[yaz]

however, there's sg strange about (renewed) bicubic. after some hard days i've installed rc4 & just for testing, i reloaded 1 older script working perfectly with rc1/2(!). nothing special: mpeg2dec3 / crop / resize(x,y,0,0.66), the same script, the same source, the same dll as before.
but now it gives some very weird artifacts. i mean, at high contrast edges color 'flashes' heavily. i've read about 'color-bleeding' but it's sg more intensive. say, white letters on black background -> there are heavy 'flashlights' betw the edges (as they were charged electrically).
i'm sure its bicubic cus when 0.66 is being lowered the artifacts get less&less intensive & below 0.5 they disapper. i know, bicubic tends to ringing above 0.5 but it's much much more.
it's not the noise cus even the heaviest filtering can't stopt it. any idea?

thx
y

ps sorry for my crap english but i'm unfamiliar with the 'terminus technicus' pretty much.