The number appended to export files should be a good starting point.

See 'pants':
void SHMProcess::getMemRanges( vector<t_memrange> & ranges )
{
t_memrange temp;
uint32_t base = d->memdescriptor->getBase();
temp.start = base + 0x1000; // more fakery.
temp.end = base + readDWord(base+readDWord(base+0x3C)+0x50)-1; // yay for magic.
temp.read = 1;
temp.write = 1;
temp.execute = 0; // fake
strcpy(temp.name,"pants");
ranges.push_back(temp);
}

In an XML file will be patches for different versions of DF and different purposes. Tools will be able to enumerate and apply the patches.

This can use hexsearch and possibly some minimal GUI (just plain terminal app or ncurses). Someone could then write the patches or apps that use them. Converting the other apps that already exist is also a possibility - embark anywhere, setting number of dwarfs on embark, nano fort, stuff like that.

For Arch Linux users, dfhack is packaged in the AUR, and has binaries available in the Arch-Games project

What the title says.

For things like telnet fortress.

Search string (while in main menu):
20 00 00 00 20 00 00 00 20 00 00 00 20 00 00 00 20 00 00 00

Find ptr to addr from search result.

Let's use those search strings and black magic :)

there must be some tag on a tile if stuff can grow on it. this has to be removed when building an obsidian "wall" (this seems to happen when a real wall is built).

i used the tool to stop an aquifer, but the walls will break soon

See: http://www.bay12forums.com/smf/index.php?topic=58809.msg1465738#msg1465738

Give the SHM a 'debug' mode, where all memory accesses are checked against the real layout of DF's memory. Encourage devs to use it for debugging. It's probably impossible to rule out all possiblilities for failure, but getting a bit closer can't hurt.
Using the 'debug' mode will obviously lead to some substantial performance hit. The user should be notified when he uses such a 'debug' mode SHM library.

WTTS

What the title says. Make it vary the message size. Both message numbers and and amount of transferred data should be tracked.

Basically:

1. client memcpys a message to the buffer
2. server memcpys it out and back in.
3. client reads the message and verifies that it's the same as the one sent in the first step
4. repeat to filter out noise.

Then we can determine the impact of other running processes, compare single core vs. multi core, etc.

See http://www.cmake.org/pipermail/cmake/2006-August/010794.html for an example.

msvcr90.dll and msvcp90.dll

would be nice to have them bundled in downloads.

http://www.bay12games.com/forum/index.php?topic=49082.0

IMHO very unlikely unless someone experienced in making Java bindings for C++ steps in.

Veins have a vptr as the first DWORD. This should be used to detect their type.

This is best seen on maps with a glacier or just about any ice. Aside from the standard vein, there's a another one that stores tile IDs for tiles that have been replaced by ice walls (guessing here, but it's probably not far from the truth). When the ice melts, the ice wall ID is then replaced by the content of this 'tile ID vein'.

See http://www.bay12forums.com/smf/index.php?topic=41916.msg1193761#msg1193761

the first coord you put in, the fact that you are in designation mode

``it automatically dumps a column of 7/7 magma on any elves on map''

I approve.

It should be able to CREATE a Context.
This means spawning a DF instance as specified by the 'user'. Provide certain degree of automation -- loading of fortresses, adventure mode, pushing input.

Also: forceful Process/Context termination and management of saves.

Another thing could be the ability to specify the type of Contexts/Processes to prefer. Say, prefer normal Processes over SHM processes. This could be good for debugging, because normal Processes can't make DF segfault from accessing invalid memory.

So, trying to write a tool to walk the list of creatures, all 990 of them- in my save, anyway. There seem to be a lot of problems, but I guess I'll create issues for them one-at-a-time. My guess with this one is that, when there isn't a custom profession, a null pointer gets set, but readSTLString doesn't check the offset it finds, it just tries to read.

GNU gdb (GDB) 7.1-debian
Copyright (C) 2010 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "i486-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
Reading symbols from /home/pdbogen/tmp/dfhack/output/dfdd...done.
(gdb) break /home/pdbogen/tmp/dfhack/library/DFProcess-linux.cpp:557 if offset==0
No source file named /home/pdbogen/tmp/dfhack/library/DFProcess-linux.cpp.
Make breakpoint pending on future shared library load? (y or [n]) y
Breakpoint 1 (/home/pdbogen/tmp/dfhack/library/DFProcess-linux.cpp:557 if offset==0) pending.
(gdb) run
Starting program: /home/pdbogen/tmp/dfhack/output/dfdd 
[Thread debugging using libthread_db enabled]
990 creatures, apparently.

Breakpoint 1, DFHack::NormalProcess::readSTLString (this=0x8098880, offset=0, buffer=0xbfffe4d1 "", bufcapacity=128) at /home/pdbogen/tmp/dfhack/library/DFProcess-linux.cpp:557
557     read(offset - sizeof(_Rep_base),sizeof(_Rep_base),(uint8_t *)&header);
(gdb) print offset
$1 = 0
(gdb) continue
Continuing.
pread failed: can't read 0xc bytes at address 0xfffffff4
errno: 16
terminate called after throwing an instance of 'DFHack::Error::MemoryAccessDenied'
  what():  SHM ACCESS DENIED

Program received signal SIGABRT, Aborted.
0xb7fe2424 in __kernel_vsyscall ()
(gdb) 

uint32_t cr_color = contentLoader.Mats->raceEx.at(b->creature->race).castes.at(b->creature->caste).ColorModifier.at(j).colorlist.at(b->creature->color[j]);
this line of code sometimes crashes when b->creature->color[j] returns more than 0, while the vecrtor is empty.

i'd really like to remove that, so i can i.e. run a script that calls dfcleanmap in a loop...
i dont think thats needed at all, it's like those annoying "do you really want to quit [Yes/no]" nag screens...
what do you think ?

This is multiple issues in one.

1. Add support for grouping of related addresses/offsets/values in the Entries -- DONE
2. Add a Meta entry that normal entries are based upon. This contains mostly meta-data for the actual entries. There's a tendency towards this already, so let's make it official. Addresses, Offsets and Hexvalues should have descriptions. -- DONE
3. Create a GUI tool to make working with the Memory.xml file(s) effective.
4. All tags should be able to have dependencies on other tags.
   All tags should be able to use hexsearch strings. Those should be used mostly in meta entries.
   For example:

1. where are they facing?
   • windmills (rotate to face the wind)
   • pumps
   • horizontal axles
   • water wheels

Use the joystick function to get it synced with DF's simulation thread.

Currently, a value of 2 for naked_designation.liquid_character indicates that water touching this tile is (or will be) salty, and 0 indicates fresh water. I have not seen 3 or 1 in the wild. I last checked this in 0.31.12.

The reveal utility does not work at all, either to reveal or unreveal. dfvdig does not recognize any tile as being a vein.
however the prospector works consistantly

ive tried to redownload 3 times and it it seems that:

downloaded tool pack: dfattachtest, dfcleanmap, dfexpbench, dfflows, dfliquids, dfprobe, dfprospector, dfreveal, dfsuspend, dfunstuck, dfvdig.

html page only has the following: dfcleanmap, dfexpbench, dfflows, dfliquids, dfposition, dfprospector, dfreveal, dfsuspend, dfunstuck, dfvdig

download page at bay12forums: dfattachtest, dfcleanmap, dfexpbench, dfflows, dfposition, dfprobe, dfprospector, dfreveal, dfsuspend, dfunstuck, dfvdig, dfdoffsets

tool pack missing: dfposition, dfdoffsets

html page missing: dfattachtest, dfoffsets

download page at bay12forums missing: dfliquids

why did that happen?

The way to handle DF crashes, invalid memory accesses and the like should be well defined.
Also define how each error and change of state of the Process class affects the modules and vice versa.

Communicate the errors properly to the client app.

Prevent segfaults from using bad modules. Possibly: On failure, bad modules are marked as such. Use smart pointers for cleanup?

All tags should be able to have dependencies on other tags.
All tags should be able to use hexsearch strings. Those should be used mostly in meta entries.
For example:

Or possibly:
[20:42:28] <jj``> the 4 should be more like

Why is it failing?

cannot be unstucked even.
the wine version works well, however.

Linux (testing 40d16)

dfdigger always designates 10 tiles and ignores the input
dfitemdesignator has problems with some item types:
Items with badly assigned materials:
item_corpsepiece : 1
item_fish : 1
item_remains : 4

Windows

dfmaterial test seems to be locking up without a loaded map
dfincremental crashes?
SDL.dll shouldn't be built with mingw and debug builds

I'm using dfhack and as soon as I start dfweather, the game starts flashing the text REC near the top of the window, and I lose my ability to interact with the game using my keyboard.

See: http://www.bay12forums.com/smf/index.php?topic=66782

"This is the Doom lever. Do nut pull."

I get an error trying to use tools that read the map.

I've tried to use dfvdig, dfreveal, dfprospector and they all output:

$ ./dfvdig
pread failed: can't read 0x30 bytes at address 0xa87f1178
errno: 16
terminate called after throwing an instance of 'DFHack::Error::MemoryAccessDenied'
what(): SHM ACCESS DENIED
Aborted

Attempted using the 0.5.0.0 release as well as the latest in git.

I'm on openSuSE 11.3 using DF 0.31.13

hexsearch needs to be able to accept a search string instead of its current input so that we don't have to hard-code the searches.

Basically a binary regexp.

Currently we don't support this version at all, which is strange, because we have every other version up to d17. From what I can tell most of the offsets are the same as d11 on windows and linux, so it shouldn't be difficult to add. That way we would have all versions from d9 on.

The old creature module needs updating to support the new DF2010 creatures.
Reflect the other modules in the SHM, extend maps and creatures to allow passing multiple objects in one message, whole vectors possibly.

See: http://gaming.stackexchange.com/questions/3435/dwarf-fortress-clutter-overview

Maybe :)

POIs

Right now we address vectors and strings based on the 'start' of the object, so we include 4 bytes for the allocator. It might be better to just address them starting at the actual data, the pointer to the start of the vector, or the string data directly. As it is now I have to remember to subtract 4 from all the addresses, then DFHack just adds the 4 back in. It would take some editing of the current addresses, but if it is acceptable to you guys I will do it, I think it would save us some hassle, as well as invalid addresses because I forgot to subtract 4 :)

See: http://wiki.osdev.org/ELF

Use the headers and possibly a small sample of .text for the md5 instead of the whole file. Make a tool that can calculate the new md5.

Is there some existing work related to this?

no usleep() in msvc.

So far, I've eliminated the g_pProcess global, so it isn't blocking anything. More refactors and design changes are required though.

Right now, the list of selected segments is determined at the start of the program. Using the feature of replacing the DF instance between searches can lead to crashes if a segment of uncertain length and position is used (like the heap).