The unit test does not really pass for now (it never did before, because it is not finished yet and used for testing the different transfer function behavior only)
Most scancodes are two hex digits long, e.g. "$24" (J). However, there
are also a couple keys like Backspace ("$e") and Escape ("$1") with a
single-digit scancode. These previously weren't restored correctly.
The matching had a few issues:
Items with the same symbol were supposed to be stacked, but the "symbol" changed from ID to object at some point. So same-type objects were no longer correctly matched.
Items with different text (e.g. amount) but everything else the same should be stacked. But simple proplist equality is obviously not sufficient to determine whether the contents of the proplist changed.
The order of the items is generally more stable now. I hope this doesn't introduce other issues.
It was no longer used. It's also just a stretched dot. Consider using an actual stretched dot instead (i.e. round GFX and Stretch = XYZ).
The sprite was also a little unclean.
Yes, someone just had the "slow-mo" switch on. I also replaced the usage of the "RaindropSmall" particle (which was a stretched dot) with an actual stretched dot.
This does not change behavior at all, as nil-objects were later ignored in the check. Now the check is not even executed.
This does make it more likely that refactoring will keep the ignoring behaviour in place, though.
Loosely related to PR #61 - has nothing to do with the solution, though.
Improved the barrel fuse effect in two ways:
1) The fuse is not at the center of the poweder keg anymore, making it more prominent
2) When carried by a clonk the effect is at the keg instead of the clonk center
I'm not sure how exactly that issue happens, but in the attached video
it looks like C4D_StaticBack was accidentally set again. As the XOR
operation can do that, I'm changing it to always set the C4D_StaticBack
bit to 0.
The old code did some obviously wrong things like calling functions
expecting relative coordinates with GetX()/GetY(). It also sometimes ran
into an infinite loop in Cloud->Place(). I fixed those issues and
cleaned up the rest to be nicer, hopefully without changing
functionality.
The previous behaviour of only checking the background broke existing
maps and triggered some (performance?) bug in the mass mover. It is
still available by setting AutoScanSideOpen=2, for symmetry with
Top/BottomOpen.
In a text line containing <c ff0000>markup</c>, C4LogBuf would
occasionally split the line after <c, producing broken markup.
Additonally, markup spanning multiple lines would always stop at the end
of the first line.
With this commit, lines are never broken within tags and any unclosed
tags are repeated on the following line.
- Each packet has a version field.
- Clients connecting to the netpuncher always send a request packet.
This allows the netpuncher to react differently depending on the
client's version.
- Encode packets as binary instead of ASCII. This allows adding fields
while maintaining compatibility.
With my fix from 9eb2478b2 ("C4NetIOUDP: Fix timeout during non-MC
connection"), C4NetIOUDP would send data packets to the master server
before the ConnOK packet. The server would then discard these data
packets, resulting in a one-second delay until retransmission.
Mark