The proposal to merge lp:~widelands-dev/widelands/new-shipping into
lp:widelands has been updated.
Status: Needs review => Work in progress
For more details, see:
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
--
Your team Widelands Developers is subscri
I have added some comments
Diff comments:
>
> === modified file 'src/logic/map_objects/tribes/ship.cc'
> --- src/logic/map_objects/tribes/ship.cc 2019-05-27 14:28:34 +
> +++ src/logic/map_objects/tribes/ship.cc 2019-09-03 15:47:16 +
> @@ -736,22 +736,215 @@
> }
> }
>
The detour-baselength assert will definitely not give any further trouble now –
I removed it ;)
Making a detour really can shorten the ship´s path by a few fields in some
cases… :P
The ship.get_nritems() assert is also fixed now, though the savegame might
remain broken.
--
I now got
Assertion failed: (ship.get_nritems() == 0), function ship_arrived,
file ../src/economy/portdock.cc, line 383.
after
./widelands --loadgame=~/Downloads/ShipCrash5.wgf
Somwhat after 1:56... from https://www.magentacloud.de/share/tu4ayusx.k
Do frisians have very special wares?
--
got another detour assert:
/widelands-repo/new-shipping/src/economy/fleet.cc:890: void
Widelands::Fleet::check_push_destination(Widelands::Game&, Widelands::Ship&,
const Widelands::PortDock&, Widelands::PortDock&, uint32_t): Zusicherung
»detour >= base_length« nicht erfüllt.
Abgebrochen
Continuous integration builds have changed state:
Travis build 5356. State: passed. Details:
https://travis-ci.org/widelands/widelands/builds/578276856.
Appveyor build 5126. State: success. Details:
The "detour >= base_length" assert should finally be fixed. Also changed the
double-colonization check as suggested.
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is subscribed to branch
lp:~widelands-dev/widelands/new-shipping.
I had a look at the check for disallowing double colonization. I think it would
be sufficient to check whether there's a building there, we don't care if it's
a port or a construction site. It might even be safer, because the enemy might
also have been faster with building something there.
--
Can you please merge trunk? I am struggling with bug 1839740 ... which is
solved in trunk.
I get the same crash:
/widelands-repo/new-shipping/src/economy/fleet.cc:888: void
Widelands::Fleet::check_push_destination(Widelands::Game&, Widelands::Ship&,
const Widelands::PortDock&,
Continuous integration builds have changed state:
Travis build 5347. State: failed. Details:
https://travis-ci.org/widelands/widelands/builds/577456819.
Appveyor build 5117. State: success. Details:
Review: Needs Fixing
OK I have a ShipCrash4.wgf at https://www.magentacloud.de/share/tu4ayusx.k
I start this with $ ./widelands --loadgame=~/Downloads/ShipCrash4.wgf
and wait about 10 Minutes for
Assertion failed: (detour >= base_length), function check_push_destination,
file
No wonder it was buggy, I used the wrong variable for calculating the length of
the detour :)
Please check whether it still happens…
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is subscribed to branch
Now I was kiled here:
Assertion failed: (detour >= base_length), function check_push_destination,
file ../src/economy/fleet.cc, line 887.
5 widelands 0x0001102aedf9
Widelands::Fleet::check_push_destination(Widelands::Game&, Widelands::Ship&,
OK, I can load the Savegame again, lets continue hunting ...
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is subscribed to branch
lp:~widelands-dev/widelands/new-shipping.
___
Mailing
The OPtr is just a wrapper for a MapObject. I use OPtr rather than
PortDock* to avoid segfaults when the portdock is destroyed. The == comparison
compares serials, so it´s 100% safe to use.
I found a potential cause for this assert in the way superfluous or invalid
destinations are discarded,
Next one please check: https://www.magentacloud.de/share/tu4ayusx.k
ShioCrash2.wgf
Crashes here:
Assertion failed: (old_index < nr_dests), function reorder_destinations,
file ../src/logic/map_objects/tribes/ship.cc, line 902.
With a Debugger I found...
Via the ReplayWriter
I believe this is a bug in trunk since I didn´t touch that code, but I´ll fix
it here. Once the portspace becomes unavailable, other ships shouldn´t be
allowed to start colonizing. No savegame needed, should be easy to reproduce.
I´ll also add this case to the testsuite.
Still wasn´t able to
OK, found an obvious Bug:
* Two ships find the same PortSpace
* Tell both of them to build a port.
4: Colonization error: unloading wares to constructionsite of atlanteans_port
(owner 4) failed.
Wares unloaded to the site: 3, max capacity: 3, remaining to unload: 13
No free capacity to unload
Hello benedikt, I am back and will try
more with a debugger attached, any news from your siede?
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is subscribed to branch
lp:~widelands-dev/widelands/new-shipping.
Mhh, it took about 15-30 minutes to trigger the asserts.
I will be "on call duty" this week, so I have even less time.
I tried one of the new features of trunk : order Soldiers to be sent
to the port immediateley once it was build.
As of some Towers "across the Sea", I was able to plan
I tried out all four UnrealAssert savegames now and *none* of them triggered an
assert fail. Do you have replays that end with assert-crashes?
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is subscribed to branch
Review: Needs Fixing
Debugging can be FUN!, well not for you Benedikt, sorry:
5: Ullr: exploration - breaking for free sea, dir=1
TrainingSite::drop_stalled_soldiers: Kicking somebody out.
Assertion failed: (detour >= base_length), function check_push_destination,
file ../src/economy/fleet.cc,
Review: Needs Fixing
OK here comes thee next (me conquering the Imperial Isaland East of mine)
5: (126x 23) expedition spot scoring, colony_scan_area == 21
5: Ullr at 2x 23: PORTSPACE found, we valued it: 0
5: Ullr at 2x 23: explore uphold, visited first time
5: Ullr: continue island
Look like
a) The attached game on partially reproduces the Issue (just slow as hell)
b) A Millitarysite near a Port was under heavy Attack, you ca find two
barbarian Inhabitat Atlanter Buidlign, ooks I lost them just during the
crash.
I set Autosae to Minutes ant will continue ... later.
--
Ouch. This means that you have an item in your portdock that doesn´t know where
it wants to go. Happens all the time in trunk, shouldn´t happen anymore in this
branch.
Did you (or some AI) destroy a port shortly before you got the error? Or
connected some previously unconnected ports on land?
OK, now its getting interesting.
My "Unreal Water" Map got totally slow and stopped with
1: attacking site at 56x 40, score 5, with 2 soldiers, attacking 2 times,
after 102 seconds
Cmd_EnemyFlagAction::execute player(1): flag->owner(3) number=2
Cmd_EnemyFlagAction::execute player(3):
Continuous integration builds have changed state:
Travis build 5332. State: passed. Details:
https://travis-ci.org/widelands/widelands/builds/572886798.
Appveyor build 5104. State: success. Details:
Just started the Debugger and wondered where this this OPtr thing comes from.
But whatever you did looks like you fixed it.
Will go on laying my Map now until I got whatever was possible.
AND thanks!
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team
Another guess pushed, this one may fix it. (untested, can´t compile atm.)
Please test…
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is subscribed to branch
lp:~widelands-dev/widelands/new-shipping.
I added the complete output to:
https://www.magentacloud.de/share/tu4ayusx.k#$/
Crash.txt
Will try a debugger now
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is subscribed to branch
Continuous integration builds have changed state:
Travis build 5331. State: failed. Details:
https://travis-ci.org/widelands/widelands/builds/572672291.
Appveyor build 5103. State: success. Details:
I added some debug output, please try again…
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is subscribed to branch
lp:~widelands-dev/widelands/new-shipping.
___
Mailing list:
Sorry still got it
SaveHandler::save_game() took 2536ms
Reloading the game from replay
=
==99248==ERROR: AddressSanitizer: heap-use-after-free on address 0x603000c71568
at pc 0x0001102296de bp 0x7ffee2224ef0 sp 0x7ffee2224ee8
I have pushed a blind guess what might be causing this, please test whether
this fixes it
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is subscribed to branch
lp:~widelands-dev/widelands/new-shipping.
Mhh, I have:
Check for working CXX compiler:
/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/c++
-- works
Apple LLVM version 10.0.1 (clang-1001.0.46.4)
And I get still get this crash at bzr9140[new-shipping]
directy when loading that file.
We had such
Whatever you did, looks like it fixed the regression tests, too.
I have about no time this weeek. But this deserves some playtesting, mmh.
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is subscribed to branch
Yes, with ASan. It complains about the usual memory leak when closing
Widelands, nothing else.
Travis is happy now, regression tests are working there.
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is subscribed to branch
Continuous integration builds have changed state:
Travis build 5317. State: passed. Details:
https://travis-ci.org/widelands/widelands/builds/570717296.
Appveyor build 5089. State: success. Details:
Continuous integration builds have changed state:
Travis build 5314. State: failed. Details:
https://travis-ci.org/widelands/widelands/builds/570532677.
Appveyor build 5087. State: success. Details:
Can´t reproduce. I downloaded your savegame, loaded it, let it run for ten
minutes gametime and then closed it, all without problems…
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is subscribed to branch
Loading the Savegame Unreal.wgf from
https://www.magentacloud.de/share/tu4ayusx.k
provokes this ASAN error in
==1418==ERROR: AddressSanitizer: heap-use-after-free on address 0x6040025b5898
at pc 0x00010b77478e ...
#2 0x10b75f8cc in Widelands::PortDock::cleanup(Widelands::EditorGameBase&)
Ah ok :) you meant, don´t use realtime as a limit for calculation steps. OK, if
playtesting shows that such a limit is necessary (atm I don´t think it is) the
limit will be implemented by max number of destinations or perhaps max total
route length.
--
> be carefull not to use (local) timings ...
For some rendering code you might use an approach
long ts = current_time() + kMaxRenderingMillis;
do {
refine_textures(...)
} while (ts < current_time());
With widelands we must not do such things, got the idea?
--
Thanks for reviewing :)
Implemented or replied to your comments.
I´ve been playing so far on Dust In The Wind where I didn´t notice performance
problems (owning ~ half the islands). A big map with many many ports would be
great for testing though.
This is sort of like the TSP with additional
Review: Needs Fixing regression tests
We have a lot of regressiontests about seafaring (for a reason :-)
./regression_test.py -b ./widelands
FAILs
test/maps/ship_transportation.wmf/scripting/test_rip_ports_with_worker_in_transit.lua
...
for (const auto& pair : dq) {
> + OPtr optr(pair.first);
> + uint32_t priority = pair.second;
> + size_t old_index = 0;
> + for (;; ++old_index) {
> + assert(old_index < nr_de
to review the proposed merge of
lp:~widelands-dev/widelands/new-shipping into lp:widelands.
___
Mailing list: https://launchpad.net/~widelands-dev
Post to : widelands-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~widelands-dev
More help
/_widelands_dev_widelands_new_shipping-5082.
--
https://code.launchpad.net/~widelands-dev/widelands/new-shipping/+merge/371155
Your team Widelands Developers is requested to review the proposed merge of
lp:~widelands-dev/widelands/new-shipping into lp:widelands.
___
Mailing list: https
wever, due to the linked bug,
trunk savegames may show irrational behaviour.
--
Your team Widelands Developers is requested to review the proposed merge of
lp:~widelands-dev/widelands/new-shipping into lp:widelands.
=== modified file 'src/economy/fleet.cc'
--- src/economy/fleet.cc 2019-04-24 15:1
49 matches
Mail list logo
