Column Misalignment On Freeze Column

[aderosso]

When we freeze a column with a large amount of data in the grid, we're seeing column misalignment.



We're able to reproduce in this showcase: https://smartclient.com/smartgwt/sho...ng_mass_update

Here's a video of reproducing the issue:

[Isomorphic]

Thanks for the video. We're on it.

We note that you saved all the records before scrolling around and doing another freeze. At that point there should be nothing special about the Mass Update sample, so our first instinct is to go try to repro this in a different sample that already has a lot of records.

We just wanted to check if you already tried that and the Mass Update sample seemed to be required for some reason?

[aderosso]

I don't think it was that sample specifically. We were able to reproduce it on:

https://smartclient.com/smartgwtee/s...amicProperties
https://smartclient.com/smartgwtee/s...e/#mass_update
https://smartclient.com/smartgwt/sho...ng_mass_update

as well as grids in our application where there are lots of records.

[Isomorphic]

Interesting - we had just concluded it won't happen without Mass Update, and we do have it reproducing on Mass Update, with a little script that automates your test setup (adds the records programmatically).

Although, "reproduce" is perhaps overstating it. We see a 3-4px offset, much smaller than yours:



We can't even get this small offset Windows 10 and FF v139, Chrome v137 and Edge v137, FF v136 and Chrome v134 on Ubuntu Linux.

We can only get this small offset on Windows10 with FF and Edge. And it's small enough to attribute to those browser's known issues with cumulative rounding errors.

Your screenshot looked like Chrome. What platform?

Also, since both teams are testing the Showcase, can you please disable any browser extensions, restart the browser and retest?

[Isomorphic]

Also, are you zoomed in Chrome? Do you know that zoom is necessarily imperfect in that browser? We explain the whole sad story here, with links to the known bugs which have not been addressed:

https://smartclient.com/smartclient-...p..browserZoom

[aderosso]

I'm checking with the developer that created the video. I'm assuming some OS scaling issue. I'm not able to reproduce on my PC without changing browser zoom or OS display scaling.

[aderosso]

We're still checking on this one. It is much more difficult to reproduce without any browser or OS scaling, but we are still able to get it to happen in our application. We're checking through a grid field by field to determine what is throwing it off.

[Isomorphic]

In your app, without room/scaling, how big of an offset can you reproduce?

Just to set expectations, if it's only around 2-3px, that's going to be due to subpixel rounding errors where there are fundamental browser bugs in the way. Basically we are providing two HTML tables to the browser (frozen and unfrozen areas), and we are telling the browser the exact pixel sizes we want, and we are not using any fractional pixels, and it still gets it wrong.

You might think that we could just detect the final size post-render and add a corrective offset, but the browser also makes subpixel rounding errors when reporting on-screen coordinates so 🤷♂ it still won't be perfect, but using this technique we could probably get the offset down to 1px.

If we added that and you tested with that in place, we would know whether you are seeing cumulative sub-pixel errors or something else entirely.