The change allows correct and optimized incremental builds with VC and
ninja but unix makefile backend will rebuild everything if anything
changes in structures.
ninja has a single build file which avoided issues if multiple targets
depend on same files. But Unix Makefiles generator user recursive make
which requires each ADD_CUSTOM_COMMAND to have only one target depending
on them.
Then makefile generator also has stupid rule that it touches all
secundary output files if primary file has been updated.
It was surprising hard to find a version that actually works correctly
for both issues. Solution is using BYPRODUCTS and refactoring command
and target dependencies.
As a bonus this change now allows build to work from source tarball if
the tarball includes git-describe.h.
Using add_custom_command allows cmake to track dependency from command
to git-describe.h that is implicit dependency of DFHackVersion.cpp.
The change also fixes issues if there is multiple build directories and
git-describe.h missing but git-dsecribe.h.tmp is present.
_exit seems to run dll unloading code which calls static destructors.
Standrd requires std::_Exit not to call destructors which makes using it
attractive in case MSVC actually follows the standard.
I'm not sure if calling GetModuleHandle in static construction is safe.
But I assumme it is and works correctly.
There is still potential issue that documentation can be understood
meaning that the HMODULE will resolve only symbols from exe while
RTLD_DEFAULT resolves all global symbols.
plen+cooked_cursor==cols => begin = -1 which is passed to substr. The
sign is incorrect as code should be removing a character from begin
instead of trying to add a character.
The runtime debug print filtering support dynamic debug print selection.
Tis patch only implements basic core support for filtering. The commands
to change the runtime filtering settings will be added in a following
patch.
But even with only this one can change filtering settings by editing
memory using a debugger. It can even be automated by using gdb break
point commands.
- Use port from remote-server.json in dfhack-run
- Make DFHACK_RUN environment variable take priority over remote-server.json
- Log current port to stderr
I often want to see multiple items quickly when trying to figure out
what states actually matter to an issue that I debug. I decided to make
it easier to quickly dump df structures with substructures and
containers. It will generate large amount of data which can be sometimes
slow to process manually. But processing can be automated using
dfhack-run lua ^<df data to inspect> and pipe to other tools (eg grep,
sed, perl, sort, uniq etc)
The bools could use acquire&release memory order or even relaxed but I
didn't think code was worth auditing for such low level optimizations.
Sequantial consistent is fast enough but much harder to use incorrectly.
The timeLast is protected by CoreSuspender lock. plugin_update is only
called when CoreSuspender lock is held.
The last_menu is protected by trackmenu_flg loads and stores.
I noticed an extra null when trying to grep dfhack.run ls output.
std::string manages null byte at end(). Pre C++11 didn't require null
termination for std::string but C++11 add the null termination to make
c_str() and data() truely const. It is undefined behavior to modify the
internal null termination but this modification sets the null to null.
The old CoreSuspender requires processing from Core::Update to allow
commands execute. But that causes issues if Core::Shutdown wants
quarentee cleanup order with std:🧵:join. Fixing shutdown ordering
adds too many branches to already fairly complex code.
I decided to try to refactor CoreSuspender to use simpler locking
locking using a std::recusive_muted as primary synchronization
primitive.
To help control when Core::Update unlocks the primary mutex there is
std::contition_variable_any and std::atomic<size_t> queue lenght
counter.
The last state variable is std::atomic<std:🧵:id> that is used to
keep track of owner thread for Core::IsSuspended query.
This should be merged only just after a release to make sure that it
gets maximum testing in develop branch before next release.
Fixes#1066
Console::lineedit can return -1 to indicate input error and -2 to
indicate the program is closing. But most users don't check possible
unusual return values which can lead to exit hang.
I noticed that multibyte characters can mess up the console state
variables. I decided to add a minimal multibyte support to make sure the
input only collects complete valid multibyte characters in case user
enters them to console.
This change assumes that UTF-32 has one to one mapping between printed
characters and char32 indexes. But I remember reading there is a few
rare corner cases with accents where character might require multiple
4byte characters too. But this patch at least changes correct handling
from about 100 characters to 99% of unicode characters.
There is a minor chance that console or init thread would access already
freed memory when core is shutting down and cleaning up state. To avoid
any danger of having random bugs caused by the potential data race I
decided to make sure the shutdown code waits for the thread to exit
first.
Windows change is completely untested. It is purely based on msdn
documentation.
I noticed that tthread is missing some c++11 features that make thread
handling code a bit easier. To be able to use those features I decided
to convert Core.cpp to use equivalent standard classes.
This patch has no functional changes.
command-prompt viewscreen may affect command execution if they are
looking for UI state. To make commands execute simillary to hotkeys or
console commands the viewscreen needs to removed from the top position.
Fixes#1194
The Screen::show takes ownership of the screen pointer. I decided to
switch the parameter to std::unique_ptr to make the pointer ownership
explicit. The unique_ptr then provides automatic screen destruction in
Screen::show unless pointer is inserted or is already in the linked list
that is managed by df.
G++ generates structure debug symbols for a few df namespace classes to
generated stub source files. I decided to test how much symbols from
those files would increase binary size. When the result was about double
size I decided to add cmake configuration option to let user easily
select if they prefer complete symbols or reduced size.
The map_block->designation.{dig,smooth} are reset to zeros when a job
posting is created for the designation. The job is then used to override
the designation state in the map_block. To make the new designation set
propogate to jobs the job structure would require updating. The update
would be possible a complex operation. The simple alternative is to
remove the job and let df create a new job in the next tick.
Fixes#1229
This makes jsoncpp a submodule that can be build directly from git
sources. This changes depends/jsoncpp to depends/jsoncpp-sub to avoid
filename conflict if someone tries to use git bisect.
jsoncpp library name changes to jsoncpp_lib_static.
jsoncpp version is the latest tagged release.
gcc supports type checks for printf parameters which can catch some hard
to reproduce bugs. Possible bugs happen when the parameter value is
intepreted differently to the variable value.
Example warnings follow
../library/LuaWrapper.cpp:1011:86: warning: format ‘%llu’ expects argument
of type ‘long long unsigned int’, but argument 3 has type ‘uint64_t
{aka long unsigned int}’ [-Wformat=]
../plugins/follow.cpp:159:35: warning: format not a string literal and no
format arguments [-Wformat-security]
vsnprintf man page claims:
"If an output error is encountered, a negative value is returned."
That means we has to call vsnprintf twice at most to have whole output
written to a string. But in case of error we return an empty string.
The code also optimizes an expected common case of outputting single
line with a small stack allocated buffer. If the stack buffer is too
small then it uses std::string::resize to allocate exactly enough memory
and writes directly to std::string.
Second call could use vsprintf because memory is known to be large
enough. But I think that difference isn't detectable outside micro
benchmarks.
Changes include
* table.getn(obj) -> #obj
* Making sure string.rep gets an integer parameter
* Optimized profiling hooks (call profiler cost from factor 40 to 10)
* Specialized parameter name lookup code for c++ __index metamod calls
* Collect source lines in time sampling variant
* Simplified prevent to always filter all children
Pepperfish Profiler can produce time sampled profiles and call entry
exit profiles. Code is verbatim copy from the lua wiki [1]. This commit
won't work alone but it exists to give author credit correctly to
Daniel.
[1] http://lua-users.org/wiki/PepperfishProfiler
Authors:
Daniel Silverstone <dsilvers@pepperfish.net>
Tom Spilman <tom@sickheadgames.com>
Ben Wilhelm <zorba-pepperfish@pavlovian.net>
The presence of syndrome data in unit.syndromes.active generates corresponding wound data in unit.body.wounds. This wound data acts to produce all of the syndrome's actual effects, including but not limited to flag changes, interaction abilities, body transformation and display name alterations. Wound data persists when syndrome data is cleared from unit.syndromes.active. Since syndrome-util did not touch wound data at all, the erase function was completely ineffective at actually removing syndromes.
Note that syndromes also generate a bunch of data in the historical figure information of units. I have observed that this historical data is sufficient to restore the syndrome in a unit following map reload (at least in adventure mode), so its clearance (which needs to also include any corresponding interaction effects) will have to be addressed in a future update. As is, syndrome erasure remains incomplete.
Stoyan Gaydarov
lethosor
Ben Lubar
Pauli
PatrikLundell
Daniel Silverstone
grubsteak
Lethosor
Dan Amlund
AtomicChicken
ThiagoLira
Warmist
Japa