Transport Tycoon Forums

Hi Cirdan, I've added a branch 'cirdan-blitter-scroll-fix' to https://github.com/JGRennison/OpenTTD-patches which fixes this issue.
It's caused by a signed/unsigned arithmetic issue when calculating the src point when scrolling vertically downwards (scroll_y < 0) in the Blitter_*::Surface::scroll methods.

You were fast—I was still looking at the code, wondering why the bug did not show up on my box, but it turns out that it is masked if integers and pointers are the same size. Thank you; I have merged your patch and pushed it to the repo, along with another commit to silence this warning.

cirdan wrote:I have merged your (JGR) patch and pushed it to the repo, along with another commit to silence this warning.

The bugicidal unguaranteed Win32 binaries are posted below: and for those who want bridges over stations (free-bridges.diff): Enjoy

Not on my compiling machine, will use this post as dummy to provide the compiled version l8er.
Thank you so much for (hopefully) fixing the problem
Cheers

Sure i will test a few things before posting and either post the executable or a bug report

After looking at the code, the same bug could affect the smallmap. Could you test this before posting a binary?

cirdan wrote:After looking at the code, the same bug could affect the smallmap. Could you test this before posting a binary?

I tried the smallmap and got the result in the screenshot.
The vertical stripes seem to appear when tiles are marked dirty and redrawn. When paused the background stays black.
There aren't any crashes when scrolling though.

Thanks for testing. Displaying the smallmap is the problem here, not scrolling. Even if there is no crash, random locations in memory may be being written to. I am working on a fix.

JGR wrote:

cirdan wrote:After looking at the code, the same bug could affect the smallmap. Could you test this before posting a binary?

I tried the smallmap and got the result in the screenshot.
The vertical stripes seem to appear when tiles are marked dirty and redrawn. When paused the background stays black.
There aren't any crashes when scrolling though.

With my Win32 binaries I saw none of this. All was as expected.

wallyweb wrote:With my Win32 binaries I saw none of this. All was as expected.

Yes, this is normal. The bug only affects 64-bit platforms, where pointers are usually wider than integers.

As JGR has posted I got the same result on Minimap. Its black with strips of the normal minimap at random points.
Even tried to enabe graphs (never used before) and other windows.
Beside the minimap it runs okay so far.
Thanks and I will wait for the fix before posting a new version of the Ubuntu binary
Cheers

The attached patch should fix the smallmap on 64-bit platforms. Can you give it a try?

cirdan wrote:The attached patch should fix the smallmap on 64-bit platforms. Can you give it a try?

All seems to work fine for me with this applied.

On an unrelated note, I get these warnings about maybe uninitialised variables when compiling:

[+] Spoiler

Code: Select all

/home/jgr/openttd/trunk4/src/elrail.cpp: In function ‘void DrawCatenaryRailway(const TileInfo*)’:
/home/jgr/openttd/trunk4/src/elrail.cpp:572:27: warning: ‘sprite_halftile’ may be used uninitialized in this function [-Wmaybe-uninitialized]
   AddSortableSpriteToDraw (ti->vd, wire_base + sss->image_offset[PCPconfig], PAL_NONE, ti->x + sss->x_offset, ti->y + sss->y_offset,

In file included from /home/jgr/openttd/trunk4/src/company_type.h:15:0,
                 from /home/jgr/openttd/trunk4/src/error.h:16,
                 from /home/jgr/openttd/trunk4/src/train_cmd.cpp:13:
/home/jgr/openttd/trunk4/src/core/enum_type.hpp: In function ‘bool TrainController(Train*, Vehicle*, bool)’:
/home/jgr/openttd/trunk4/src/core/enum_type.hpp:109:3: warning: ‘chosen_trackdir’ may be used uninitialized in this function [-Wmaybe-uninitialized]
   m_val = (storage_type)e;
   ^
/home/jgr/openttd/trunk4/src/train_cmd.cpp:3610:13: note: ‘chosen_trackdir’ was declared here
    Trackdir chosen_trackdir;

JGR wrote:All seems to work fine for me with this applied.

Thank you, I have pushed the fix to the repository.

JGR wrote:On an unrelated note, I get these warnings about maybe uninitialised variables when compiling:

Yes, I am aware of those, and they are false positives. I could silence them by initialising the relevant variables with a constant, but I am not too keen on writing worse code to appease the compiler.

cirdan wrote: I have pushed the fix to the repository.

More bugicide ...
Unguaranteed Windows 32bit binaries:
JGR's tracerestrict version is available here.

Enjoy!

Looks fine over here so I post the usual 64bit version for Ubuntu 14.04
Still see no reason to switch to 16.04 haha

Thanks JGR and Cirdan for the fast fixes.

And as well JGR tracerestriction branch.



Cheers

wallyweb wrote:JGR's tracerestrict version will be posted here when we have his push.

TrueSatan wrote:JGR tracerestriction will be posted as soon as JGR has updated the branch

This is on Github now.

JGR wrote:This is on Github now.

For those who want the JGR tracerestricted Windows 32 bit bugicide we have:
The original cirdan variety is available here.

Enjoy.

This might be something that's due to self-compiling/patching, but I have crashes when building a rail X crossing directly on a bridge head, and then removing only the crossing track. Works fine for roads and demolishing the entire bridge, though.

...

Kinda almost feel like this came up before, but, eh, I don't know. Maybe someone can double-check with the builds posted here?

Confirmed, the crash happens when trying to update the signals around the segment that contains the bridge head. The attached patch should fix this. Note that it may take some time until I can push this to the repository.

Now that's what I call great costumer service ... thanks.