OpenTTD/known-bugs.txt

374 lines
19 KiB
Plaintext
Raw Normal View History

OpenTTD's known bugs
Last updated: 2011-12-05
Release version: 1.1.4
------------------------------------------------------------------------
Table of contents
-----------------
1.0) About
2.0) Known bugs in this release
* 2.1) Known bugs that will not be solved
1.0) About
---- -----
All bugs listed below are marked as known. Please do not submit any bugs
that are the same as these. If you do, do not act surprised, because
we WILL flame you!!
Of course if you have more knowledge about any of these bugs, have more
specifics, we welcome you to report them. React to the given bug indicated
by the number below on http://bugs.openttd.org.
2.0) Known bugs in this release
---- --------------------------
2011-01-09 16:14:22 +01:00
The following bugs are known to exist in this release and we intend to
fix them. Some bugs are known but are not fixable or fixing them would
cause further problems. Those bugs can be found in the "Known bugs that
will not be solved" section.
The bugs in this section all refer to a ticket in our bug tracking system
that you can find at: http://bugs.openttd.org
If the bugs are closed but still listed here it means that the bug is fixed
and that the nightlies and next major release will not have that bug.
Issues prefixed with [OSX] are required to be fixed before we consider
officially supporting Mac OS X again. For now it remains unsupported and
we only apply bug fixes provided by the community but we are unable to fix
these bugs ourselves.
- 4744 [OSX] Crash when switching to full screen with OS X Lion
- 4689 [OSX] Crash when hiding window after coming from full screen
- 4594 [Windows] Minimizing the windows when a modal error message is
shown makes it difficult to exit the game
- 4420 [OSX] OS' mouse pointer still shows
- 2484 [OSX] Cannot enter CJK characters
2.1) Known bugs that will not be solved
---- ----------------------------------
This section lists all known bugs that we do not intend to fix and the
reasons why we think that fixing them is infeasible. We might make some
minor improvements that reduce the scope of these bugs, but we will not
be able to completely fix them.
No suitable AI can be found
If you have no AIs and an AI is started the so-called 'dummy' AI will
be loaded. This AI does nothing but writing a message on the AI debug
window and showing a red warning. There are basically two solutions
for this problem: you must change the settings so no AI is started,
this is done in the difficulty settings window. The other solution is
acquiring (downloading) some AI. The easiest way to do this is via
the "Check Online Content" button in the main (intro) menu or via
"AI Settings" -> "Select AI" -> "Check Online Content" which is also
accessed via the main menu.
After a while of playing, colours get corrupted
In Windows 7 the background slideshow corrupts the colour mapping of
OpenTTD's 8bpp screen modes. Workarounds for this are:
a) Switching to windowed mode, instead of fullscreen
b) Switching off background slideshow
c) Setting up the 32bpp-anim or 32bpp-optimized blitter
Long delay between switching songs/music
On Windows there is a delay of a (few) second(s) between switching of
songs for the "win32" driver. This delay is caused by the fact that
opening a MIDI file via MCI is extremely slow.
DirectMusic, known as "dmusic" in OpenTTD, has a much shorter delay.
However, under some circumstances DirectMusic does not reset its
state properly causing wrongly pitched/bad sounding songs. This
problem is in DirectMusic as it is reproducable with Microsoft's
DirectMusic Producer. DirectMusic has been deprecated since 2004
and as such has no support for 64 bits OpenTTD.
As a delay is favourable over bad sounding music the "win32" driver
is the default driver for OpenTTD. You can change this default by
setting the "musicdriver" in your openttd.cfg to "dmusic".
Custom vehicle type name is incorrectly aligned
Some NewGRFs use sprites that are bigger than normal in the "buy
vehicle" window. Due to this they have to encode an offset for the
vehicle type name. Upon renaming the vehicle type this encoded offset
is stripped from the name because the "edit box" cannot show this
encoding. As a result the custom vehicle type names will get the
default alignment. The only way to (partly) fix this is by adding
spaces to the custom name.
Clipping problems [FS#119]
In some cases sprites are not drawn as one would expect. Examples of
this are aircraft that might be hidden below the runway or trees that
in some cases are rendered over vehicles.
The primary cause of this problem is that OpenTTD does not have enough
data (like a 3D model) to properly determine what needs to be drawn in
front of what. OpenTTD has bounding boxes but in lots of cases they
are either too big or too small and then cause problems with what
needs to be drawn in front of what. Also some visual tricks are used.
For example trains at 8 pixels high, the catenary needs to be drawn
above that. When you want to draw bridges on top of that, which are
only one height level (= 8 pixels) higher, you are getting into some
big problems.
We can not change the height levels; it would require us to either
redraw all vehicle or all landscape graphics. Doing so would mean we
leave the Transport Tycoon graphics, which in effect means OpenTTD
will not be a Transport Tycoon clone anymore.
Mouse scrolling not possible at the edges of the screen [FS#383] [FS#3966]
Scrolling the viewport with the mouse cursor at the edges of the screen
in the same direction of the edge will fail. If the cursor is near the
edge the scrolling will be very slow.
OpenTTD only receives cursor position updates when the cursor is inside
OpenTTD's window. It is not told how far you have moved the cursor
outside of OpenTTD's window.
Lost trains ignore (block) exit signals [FS#1473]
If trains are lost they ignore block exit signals, blocking junctions
with presignals. This is caused because the path finders cannot tell
where the train needs to go. As such a random direction is chosen at
each junction. This causes the trains to occasionally to make choices
that are unwanted from a player's point of view.
This will not be fixed because lost trains are in almost all cases a
network problem, e.g. a train can never reach a specific place. This
makes the impact of fixing the bug enormously small against the
amount of work needed to write a system that prevents the lost trains
from taking the wrong direction.
Vehicle owner of last transfer leg gets paid for all [FS#2427]
When you make a transfer system that switches vehicle owners. This
is only possible with 'industry stations', e.g. the oil rig station
the owner of the vehicle that does the final delivery gets paid for
the whole trip. It is not shared amongst the different vehicle
owners that have participated in transporting the cargo.
This sharing is not done because it would enormously increase the
memory and CPU usage in big games for something that is happening
in only one corner case. We think it is not worth the effort until
sharing of stations is an official feature.
Forbid 90 degree turns does not work for crossing PBS paths [FS#2737]
When you run a train through itself on a X junction with PBS turned on
the train will not obey the 'forbid 90 degree turns' setting. This is
due to the fact that we can not be sure that the setting was turned
off when the track was reserved, which means that we assume it was
turned on and that the setting does not hold at the time. We made it
this way to allow one to change the setting in-game, but it breaks
slightly when you are running your train through itself. Running a
train through means that your network is broken and is thus a user
error which OpenTTD tries to graciously handle.
Fixing this bug means that we need to record whether this particular
setting was turned on or off at the time the reservation was made. This
means adding quite a bit of data to the savegame for solving an issue
that is basically an user error. We think it is not worth the effort.
Duplicate (station) names after renaming [FS#3204]
After renaming stations one can create duplicate station names. This
is done giving a station the same custom name as another station with
an automatically generated name.
The major part of this problem is that station names are translatable.
Meaning that a station is called e.g. '<TOWN> Central' in English and
'<TOWN> Centraal' in Dutch. This means that in network games the
renaming of a town could cause the rename to succeed on some clients
and fail at others. This creates an inconsistent game state that will
be seen as a 'desync'. Secondly the custom names are intended to fall
completely outside of the '<TOWN> <name>' naming of stations, so when
you rename a town all station names are updated accordingly.
As a result the decision has been made that all custom names are only
compared to the other custom names in the same class and not compared
to the automatically generated names.
Extreme CPU usage/hangs when using SDL and PulseAudio [FS#3294]
OpenTTD hangs/freezes when closing, OpenTTD is slow, OpenTTD uses a lot of CPU
OpenTTD can be extremely slow/use a lot of CPU when the sound is
played via SDL and then through PulseAudio's ALSA wrapper. Under the
same configuration OpenTTD, or rather SDL, might hang when exiting
the game. This problem is seen most in Ubuntu 9.04 and higher.
This is because recent versions of the PulseAudio sound server are
configured to use timer-based audio scheduling rather than
interrupt-based audio scheduling. Configuring PulseAudio to force
use of interrupt-based scheduling may resolve sound problems for
some users. Under recent versions of Ubuntu Linux (9.04 and higher)
this can be accomplished by changing the following line in the
/etc/pulse/default.pa file:
load-module module-udev-detect
to
load-module module-udev-detect tsched=0
Note that PulseAudio must be restarted for changes to take effect.
Older versions of PulseAudio may use the module-hal-detect module
instead. Adding tsched=0 to the end of that line will have a similar
effect.
Another possible solution is selecting the "pulse" backend of SDL
by either using "SDL_AUDIODRIVER=pulse openttd" at the command
prompt or installing the 'libsdl1.2debian-pulseaudio' package from
Ubuntu's Universe repository. For other distributions a similar
package needs to be installed.
OpenTTD not properly resizing with SDL on X [FS#3305]
Under some X window managers OpenTTD's window does not properly
resize. You will either end up with a black bar at the right/bottom
side of the window or you cannot see the right/bottom of the window,
e.g you cannot see the status bar. The problem is that OpenTTD does
not always receive a resize event from SDL making it impossible for
OpenTTD to know that the window was resized; sometimes moving the
window will solve the problem.
Window managers that are known to exhibit this behaviour are KDE's
and GNOME's. With the XFCE's and LXDE's window managers the resize
event is sent when the user releases the mouse.
Incorrect colours, crashes upon exit, debug warnings and smears upon
window resizing with SDL on Mac OS X [FS#3447]
Video handling with (lib)SDL under Mac OS X is known to fail on some
versions of Mac OS X with some hardware configurations. Some of the
problems happen only under some circumstances whereas others are
always present.
We suggest that the SDL video/sound backend is not used for OpenTTD
in combinations with Mac OS X.
Train crashes entering same junction from block and path signals [FS#3928]
When a train has reserved a path from a path signal to a two way
block signal and the reservation passes a path signal through the
back another train can enter the reserved path (only) via that same
two way block signal.
The reason for this has to do with optimisation; to fix this issue
the signal update has to pass all path signals until it finds either
a train or a backwards facing signal. This is a very expensive task.
The (signal) setups that allow these crashes can furthermore be
considered incorrectly signalled; one extra safe waiting point for
the train entering from path signal just after the backwards facing
signal (from the path signal train) resolves the issue.
Crashes when playing music [FS#3941]
Mac OS X's QuickTime (default music driver) and Windows' MCI (win32
music driver) crash on some songs from OpenMSX. OpenTTD cannot do
anything about this. Please report these crashes to the authors of
OpenMSX so the crash causing songs can be removed or fixed.
Crashes when run in a VM using Parallels Desktop [FS#4003]
When the Windows version of OpenTTD is executed in a VM under
Parallels Desktop a privileged instruction exception may be thrown.
As OpenTTD works natively on OSX as well as natively on Windows and
these native builds both don't exhibit this behaviour this crash is
most likely due to a bug in the virtual machine, something out of
the scope of OpenTTD. Most likely this is due to Parallels Desktop
lacking support for RDTSC calls. The problem can be avoided by using
other VM-software, Wine, or running natively on OSX.
OpenTTD hangs when started on 32 bits Windows [FS#4083]
Under some circumstances OpenTTD might hang for hours on the
initialisation of the music driver. The exact circumstances are
unknown except that it is the "dmusic" music driver that has the
problem and that the "win32" music driver does not.
As a result using the "win32" music driver will work around this
issue.
As the exact circumstances are unknown, and the obvious
configuration settings related to the music driver are at their
default we are not able to detect this failure, except when Windows'
music initialisation function returns after several hours and then
there is no point in switching the music driver anymore.
The reason we still use the "win32" music driver as default are
described in the "Long delay between switching music/song" section
of this document.
Pre- and exit signals are not dragged [FS#4378]
Unlike all other signal types, the entry- and exit signals are not
dragged but instead normal signals are placed on subsequent track
sections. This is done on purpose as this is the usually more con-
venient solution. There are little to no occasions where more than
one entry or exit signal in a row are useful. This is different
for all other signal types where several in a row can serve one
purpose or another.
Station build date is incorrect [FS#4415]
The tile query tool will show the date of the last (re)construction
at the station and not the date of the first construction. This is
due to compatability reasons with NewGRFs and the fact that it is
wrong to say that the station is built in a particular year when it
was completely destroyed/rebuilt later on.
The tile query tool can be fixed by changing the "Build date" text
to "Date at which the last (re)construction took place" but this is
deemed too specific and long for that window.
Can't change volume inside OpenTTD [FS#4416]
Some backends do not provide a means to change the volume of sound
effects or music. The mixing of music and sound is left to external
libraries/the operating system we can't handle the volume control
in OpenTTD. As a result you can't change the volume inside OpenTTD
for backends such as SDL; just use the volume control provided by
your operating system.
Can't run OpenTTD with the -d option from a MSYS console [FS#4587]
The MSYS console does not allow OpenTTD to open an extra console for
debugging output. Compiling OpenTTD with the --enable-console
configure option prevents this issue and allows the -d option to use
the MSYS console for its output.
Unreadable characters for non-latin locales [FS#4607]
OpenTTD does not ship a non-latin font in its graphics files. As a
result OpenTTD needs to acquire the font from somewhere else. What
OpenTTD does is ask the operating system, or a system library, for
the best font for a given language if the currently loaded font
does not provide all characters of the chosen translation. This
means that OpenTTD has no influence over the quality of the chosen
font; it just does the best it can do.
If the text is unreadable there are several steps that you can take
to improve this. The first step is finding a good font and configure
this in the configuration file. See section 9.0 of readme.txt for
more information. You can also increase the font size to make the
characters bigger and possible better readable.
If the problem is with the clarity of the font you might want to
enable anti-aliasing by setting the small_aa/medium_aa/large_aa
settings to "true". However, anti-aliasing only works when a 32 bits
blitter has been selected, e.g. blitter = "32bpp-anim", as with the
8 bits blitter there are not enough colours to properly perform the
anti-aliasing.
(Temporary) wrong colours when switching to full screen [FS#4511]:
On Windows it can happen that you temporarily see wrong colours
when switching to full screen OpenTTD, either by starting
OpenTTD in full screen mode, changing to full screen mode or by
ALT-TAB-ing into a full screen OpenTTD. This is caused by the
fact that OpenTTD, by default, uses 8bpp paletted output. The
wrong colours you are seeing is a temporary effect of the video
driver switching to 8bpp palette mode.
This issue can be worked around in two ways:
a) Setting fullscreen_bpp to 32
b) Setting up the 32bpp-anim or 32bpp-optimized blitter
Train does not crash with itself [FS#4635]:
When a train drives in a circle the front engine passes through
wagons of the same train without crashing. This is intentional.
Signals are only aware of tracks, they do not consider the train
length and whether there would be enough room for a train in some
circle it might drive on. Also the path a train might take is not
necessarily known when passing a signal.
Checking all circumstances would take a lot of additional computational
power for signals, which is not considered worth the effort, as
it does not add anything to gameplay.
Nevertheless trains shall not crash in normal operation, so making
a train not crash with itself is the best solution for everyone.
Vehicles not keeping their "maximum" speed [FS#4815]:
Vehicles that have not enough power to reach and maintain their
advertised maximum speed might be constantly jumping between two
speeds. This is due to the fact that speed and its calculations
are done with integral numbers instead of floating point numbers.
As a result of this a vehicle will never reach its equilibrium
between the drag/friction and propulsion. So in effect it will be
in a vicious circle of speeding up and slowing down due to being
just at the other side of the equilibrium.
Not speeding up when near the equilibrium will cause the vehicle
to never come in the neighbourhood of the equilibrium and not
slowing down when near the equilibrium will cause the vehicle
to never slow down towards the equilibrium once it has come down
a hill.
It is possible to calculate whether the equilibrium will be
passed, but then all acceleration calculations need to be done
twice.