Update of bug #18776 (project freeciv):
Status: Ready For Test = In Progress
Planned Release: 2.3.1,2.4.0 = 2.3.2,2.4.0
___
Reply to this item at:
Follow-up Comment #30, bug #18776 (project freeciv):
Last call for 2.3.1 -- I will make the release starting tomorrow
morning.
I will create a minimal version which does allow the /remove command to work.
A hint is comment #21:
It does not crash. And i tested it without your previous
Follow-up Comment #31, bug #18776 (project freeciv):
The attached two patches should fix the bug, but (see below)
0005-basic-fix-known-map-in-savegame2.patch
- do not use the player_count() but the highest used player slot index
- fast fix to be used in 2.3.1
- it only changes the function used
On 15 November 2011 16:39, Michal Mazurek
no-reply.invalid-addr...@gna.org wrote:
This part is really perplexing:
+verbose+
2782 known[l * MAP_INDEX_SIZE + tile_index(ptile)]
2783 |= (1u (p - l * 8));
-verbose-
Suppose there is a player with index 60. p = 60,
Follow-up Comment #26, bug #18776 (project freeciv):
by Marko Lindqvist:
I think someone has mixed number of bits and hex values at some
point. While we read 8 hex values to known[], here we handle it
as bits already, so that 8 should be 32 (p is always 0 to 31
bigger than l*32)
I think that
Follow-up Comment #27, bug #18776 (project freeciv):
p = player_index(pplayer);
l = player_index(pplayer) / 32;
(p - l * 32)
Can't we just use modulo here?
Won't this break existing save games? If i saved using the '8' code, I cannot
load using the '32' code. Well, at least if
Follow-up Comment #28, bug #18776 (project freeciv):
What does this code save anyway? Map tiles without fog of war? Can't this be
recreated after /load?
___
Reply to this item at:
http://gna.org/bugs/?18776
Follow-up Comment #21, bug #18776 (project freeciv):
It does not crash. And i tested it without your previous patches, the ones
with vision - while they might have fixed something they were not the cause of
my segfaults.
Still, could you help me understand: 1u (p - l * 8), which really means
Follow-up Comment #22, bug #18776 (project freeciv):
I removed half the players in LTeX (idlers). The memory usage dropped by
about half. I saved the game then restarted the server, and a turn was made -
everything is ok.
___
Reply to
Follow-up Comment #23, bug #18776 (project freeciv):
I removed half the players in LTeX (idlers). The memory usage
dropped by about half. I saved the game then restarted the
server, and a turn was made - everything is ok.
So the last patch fixed it? If this is the case, should both patches be
Follow-up Comment #24, bug #18776 (project freeciv):
[bit] = p - floor(p/32) * 8
which limits it to values between 0 and 31.
Does it? p - player index in 0..127. let p = 127.
p - floor(p/32) * 8 = 127 - 3 * 8 = 103.
I run just this part in a test c program. This is really weird.
#include
Follow-up Comment #25, bug #18776 (project freeciv):
Why isn't that 1 just shifted by p % 32?
I do not know! I did take the code from the old version, (tried to)
understand it and used it in savegame2.c - and it did work! As you say, this
seems to be at least strange ...
Follow-up Comment #19, bug #18776 (project freeciv):
The allocated table is not always of constant size:
+verbose+
int lines = player_count()/32 + 1;
---
./player.c:640: int player_count(void)
./player.c:641: {
./player.c:642: return player_slots.used_slots;
./player.c:643: }
+verbose+
Should
Follow-up Comment #20, bug #18776 (project freeciv):
Thanks for the analyses! Could you test the attached patch? It does not use
player_count() but a new function player_slot_max_used_number() which returns
the highest used player slot index. Thus, there should be enough memory be
allocated.
Follow-up Comment #18, bug #18776 (project freeciv):
any update to the status of this bug? I can not reproduce a segfault with the
patch applied. I will commit it for 2.3.1 if there are no comments.
___
Reply to this item at:
Follow-up Comment #16, bug #18776 (project freeciv):
2785 p = player_index(pplayer);
2786 l = p / 32;
2787 known[l * MAP_INDEX_SIZE + tile_index(ptile)]
2788 |= (1u (p - l * 8));
Are player indexes updated for existing players,
Follow-up Comment #17, bug #18776 (project freeciv):
Are player indexes updated for existing players, after one from
the middle is removed?
player_index() calls player_number() which calls player_slot_index(); thus,
the index of a player does not change if one is removed. Furthermore,
Follow-up Comment #14, bug #18776 (project freeciv):
I applied file #14319 and file #14320.
... /removed 15 players ...
/save removed
Segmentation fault (core dumped)
#0 0x0019925c in sg_save_map_known (saving=0x208c3b000) at
savegame2.c:2787
2787known[l *
Follow-up Comment #15, bug #18776 (project freeciv):
I did try to reproduce the error with the savegame. I did removed some
players and set others to AI. It runs a autogame without problems. So there
has to be something else. Di dyou do something else between starting the game,
calling remove
Follow-up Comment #11, bug #18776 (project freeciv):
found out after my last message that loading such a savegame is still not
possible (see comment #2). There is more to this bug ...
___
Reply to this item at:
Follow-up Comment #12, bug #18776 (project freeciv):
updated patch - remove some duplicated code
(file #14319)
___
Additional Item Attachment:
File name: 0034-fix-bug-in-player_map_free.patch Size:2 KB
Follow-up Comment #13, bug #18776 (project freeciv):
found out after my last message that loading such a savegame is
still not possible (see comment #2). There is more to this bug
...
this is fixed in bug #18886
___
Reply to this item
Update of bug #18776 (project freeciv):
Category:None = general
Status:None = Ready For Test
Assigned to:None = syntron
Planned Release:
Update of bug #18776 (project freeciv):
Planned Release: = 2.3.2,2.4.0
___
Follow-up Comment #9:
Setting 2.3.2 target, although it can go in 2.3.1 if a working patch is ready
in time.
Follow-up Comment #8, bug #18776 (project freeciv):
It still crashes, though there were no assertion failures this time. The
crash is at TC, apparently during the /save phase.
#0 0x001991bc in sg_save_map_known (saving=0x20c060c00) at
savegame2.c:2782
2782known[l *
Follow-up Comment #7, bug #18776 (project freeciv):
updated patch (it compiles; test still needed!)
(file #14287)
___
Additional Item Attachment:
File name: 0040-fix-bug-in-player_map_free.patch Size:1 KB
Follow-up Comment #5, bug #18776 (project freeciv):
Did anyone have a chance to look at this bug?
___
Reply to this item at:
http://gna.org/bugs/?18776
___
Message sent via/by Gna!
Follow-up Comment #6, bug #18776 (project freeciv):
untested possible patch (not even compile tested) ... could you test it with
your savegame? (first check if it still crashs - then test with a patched
version of the server)
(file #14245)
On Sun, 02 Oct 2011 15:07:05 +0200 Michal Mazurek wrote:
I removed around 15 players using the /remove command. The server crashed. I
only have a savegame from right before the /removes (which implies the turn
change did not get to save). Here is what i got:
#0 0x001ab454 in
Follow-up Comment #3, bug #18776 (project freeciv):
Version is S2.3
#0 0x001ab454 in calc_civ_score (pplayer=0x20d084000) at
score.c:199
199 pcmap-player[player_index(owner)].landarea++;
(gdb) bt full
#0 0x001ab454 in calc_civ_score (pplayer=0x20d084000) at
Follow-up Comment #4, bug #18776 (project freeciv):
Looks to me like game has borders and tile ownership wasn't cleared when
player was removed.
The only function that (to me) plays with map is player_map_free, and it only
removes vision_site.
BTW, if player had tiles ownership, there should
URL:
http://gna.org/bugs/?18776
Summary: /remove perhaps faulty
Project: Freeciv
Submitted by: akfaew
Submitted on: Sun Oct 2 13:07:04 2011
Category: None
Severity: 3 - Normal
Priority: 5
Follow-up Comment #1, bug #18776 (project freeciv):
I managed to recreate the crash. Again I removed the players.
... remove 15 players ...
/remove alex pancho
2: Removing player Alex Pancho.
Removed player Alex Pancho from the game.
/save test
Game saved as
Follow-up Comment #2, bug #18776 (project freeciv):
After loading test.sav.gz:
1: Player shuffle 22 used two times
1: Failure loading savegame!
The savegame cannot be loaded.
Complete log:
[15:17:ttyp9][longturn@spock:~/server/ltex2.3:11]$ sh ltex2.3.sh test.sav.gz
34 matches
Mail list logo
