michi_cc wrote:

Just a quick comment about the patch:
If I've not missed something, the handling of CFLAGS in generate_src_osx() isn't
correct. The CFLAGS for the 32-bit targets are set before the loop, and the 64-bit
flags inside the loop. If $enable_universal contains e.g "ppc64 ppc" the 64-bit flags
will be used for the ppc target.

This comment was imported from FlySpray: https://bugs.openttd.org/task/4210#comment9042

SmatZ wrote:

Can you try running openttd under valgrind, if possible?

This comment was imported from FlySpray: https://bugs.openttd.org/task/4210#comment9044

planetmaker wrote:

I've not yet much experience with valgrind and it might well be possible I'm not yet using a properly configured binary nor proper options for valgrind. A first test yields the attached result

EDIT: Attached also a screenshot which shows the initial configuration which also restored wrt the blackend area in this pattern after a complete redraw of the window e.g. when restoring to initial size from minimized state.

EDIT2:
additional summary after ending OpenTTD:

==97349==
==97349== HEAP SUMMARY:
==97349== in use at exit: 9,694,396 bytes in 22,558 blocks
==97349== total heap usage: 778,659 allocs, 756,101 frees, 744,355,257 bytes allocated
==97349==
==97349== LEAK SUMMARY:
==97349== definitely lost: 4,901 bytes in 85 blocks
==97349== indirectly lost: 684 bytes in 14 blocks
==97349== possibly lost: 7,300 bytes in 14 blocks
==97349== still reachable: 9,681,511 bytes in 22,445 blocks
==97349== suppressed: 0 bytes in 0 blocks
==97349== Rerun with --leak-check=full to see details of leaked memory
==97349==
==97349== For counts of detected and suppressed errors, rerun with: -v
==97349== Use --track-origins=yes to see where uninitialised values come from
==97349== ERROR SUMMARY: 303 errors from 7 contexts (suppressed: 1 from 1)

Attachments

• valgrind.txt (8.36 KiB)
• Bildschirmfoto 2010-11-09 um 09.16.39.png (463.68 KiB)

This comment was imported from FlySpray: https://bugs.openttd.org/task/4210#comment9047

frosch wrote:

Turned out as compiler bug:
In viewport.cpp ViewportAddLandscape() "direction ^= 1" is miscompiled.
"direction ^= true" fails as well. "direction = !direction" works. making "direction" a static variable (instead of most likely a register variable) makes it work, too.

This comment was imported from FlySpray: https://bugs.openttd.org/task/4210#comment9052

planetmaker closed the ticket.

Reason for closing: Implemented

In r21149

This comment was imported from FlySpray: https://bugs.openttd.org/task/4210