[graf2d] Fix Setting window size in batch mode

[rlalik]

The current implementation doesn't make sense at all? Was it some accidentally committed code?

@linev It was your commit? Was it mistake?

771c44692bcb graf2d/gpad/src/TCanvas.cxx (Sergey Linev      2023-07-06 16:42:42 +0200 2151)    if (fBatch && !IsWeb())
d023aecbdda4 graf2d/gpad/src/TCanvas.cxx (Sergey Linev      2020-05-04 19:06:00 +0200 2152)       SetCanvasSize((ww + fCw) / 2, (wh + fCh) / 2);
d023aecbdda4 graf2d/gpad/src/TCanvas.cxx (Sergey Linev      2020-05-04 19:06:00 +0200 2153)    else if (fCanvasImp)
d023aecbdda4 graf2d/gpad/src/TCanvas.cxx (Sergey Linev      2020-05-04 19:06:00 +0200 2154)       fCanvasImp->SetWindowSize(ww, wh);

• tested changes locally
• updated the docs (if necessary)

[linev]

@rlalik

No, this kind of calculation was there since a while. See:

https://github.com/root-project/root/blob/v6-24-00-patches/graf2d/gpad/src/TCanvas.cxx#L2201

I only add special handling for web case

[linev]

I have no idea why code looks like this but I prefer not to touch it.
SetWindowSize has no real meaning for the batch - one should use SetCanvasSize instead.

[rlalik]

Well, in the batch mode there is indeed no window but this code changes the canvas size. And I would say, for compatibility reasons, should work properly. Otherwise macro run in batch mode will produce wrongly sized canvases.

[ferdymercury]

related: #11004 and c3de92f

[ferdymercury]

@linev It was your commit? Was it mistake?

It was this commit by @couet suggested by Wile in 2017

281c670
https://root-forum.cern.ch/t/tcanvas-setwindowsize-extension-proposal/26226

According to Wile, it's because:

TCanvas shows the typical usage “c->SetWindowSize(w + (w - c->GetWw()), h + (h - c->GetWh()));”, so we need to “recalculate” parameters’ values (in order to get the requested original “w” and “h”).

[rlalik]

I do not understand the logic behind it. In the batch mode SetCanvasSize((ww + fCw) / 2, (wh + fCh) / 2); causes that the new canvas size has size of half window. I don't know how it explains the message in commit 281c670. Perhaps it worked with some other logic which was changed in the meantime?

Did Wile test his suggestion? Is there any trace in the issue or in the forum about that? Perhaps there was miscommunication and Wile's suggestion as implemented incorrectly?

[rlalik]

Ok, now I indeed understand, because c->SetWindowSize(c->GetWindowWidth() + (c->GetWindowWidth() - c->GetWw()), c->GetWindowHeight() + (c->GetWindowHeight() - c->GetWh())); the Ww() is 0 so we set double of window width and we need to divide by half later.

But you see that this function will work properly only in this specific case, someone need to call it in this particular way, otherwise it will give wrong result. This is so wrong in my opinion. The logic is flaved here and should be handled different way.

[ferdymercury]

But you see that this function will work properly only in this specific case, someone need to call it in this particular way, otherwise it will give wrong result. This is so wrong in my opinion. The logic is flaved here and should be handled different way.

Yes, I agree all this is really messy.