[PATCH] Erroneous data sent to scrollbar code.

[Josh]

Steps to reproduce:

1. Make sure you have no demo maps
2. Open the single-player gamesetup screen
3. Set the filters to show the empty list of demo maps
4. Change the filters
5. View warnings

I tried to find the source of the issues, and it seems that some piece of code is setting the area available to the scrollbar to a negative number through SetLength which should never happen.

Here are the places that call that function:

gui/CList.cpp

151: GetScrollBar(0).SetLength( rect.bottom - rect.top );

gui/COList.cpp

97: GetScrollBar(0).SetLength( rect.bottom - rect.top );

gui/CInput.cpp

719: GetScrollBar(0).SetLength(m_CachedActualSize.bottom - m_CachedActualSize.top);

749: GetScrollBar(0).SetLength(0.f);

753: GetScrollBar(0).SetLength( m_CachedActualSize.bottom - m_CachedActualSize.top );

952: GetScrollBar(0).SetLength( m_CachedActualSize.bottom - m_CachedActualSize.top );

993: GetScrollBar(0).SetLength(m_CachedActualSize.bottom - m_CachedActualSize.top);

gui/CText.cpp

123: GetScrollBar(0).SetLength(m_CachedActualSize.bottom - m_CachedActualSize.top);

179: GetScrollBar(0).SetLength( m_CachedActualSize.bottom - m_CachedActualSize.top );

Alright, it appears the problem is in gui/CList.cpp

[fabio]

This looks the same issue of #2071.

[Josh]

I think this is different, the errors here are:

WARNING: Minimum scrollbar size of 4 is larger than the total scrollbar size of -24

[historic_bruno]

This isn't reproducible for me if r13773 is reverted.

[sanderd17]

I believe Josh added the warning to show when faulty data is being send. It looks like the faulty data has been in for a very long time, but it's only visible now thanks to r13773.

[Josh]

Replying to historic_bruno:

This isn't reproducible for me if r13773 is reverted.

My changes also added some warnings if the data to some stuff was totally wrong. This was intended to help debug issues when working with the moddable GUI code, but apparently some of the existing C++ code is calculating non-sensible data that had been ignored before but now is warned about. One option would be just removing the warning and hoping the code keeps working, but that is really hacky and should be avoided. The best solution would most likely just be fixing the original bug (not even mentioning that fixing the original bug should be really easy once we identify the source).

If someone could figure out where the bad calculation originates, I would happily try to fix it.

[Josh]

Check our data validity before using SetLength.

[Josh]

Actually, on further thought, there are valid situations during normal use where the scrollbar length could be zero (Example: Make window smaller than the smallest default resolution). So I think it would be best to remove the warning.

[Josh]

Fixed in the (yet to be merged) lobby fork.

[wraitii]

Pretty sure this is fixed now since the lobby fork has been merged and I can't reproduce the issue.