migration to Codelite
Habbit
Reasons are:
- Codelite is more adopted for C++, than Eclipse which is fine for
Java
- Codelite seems faster
- Codelite has GUI with many features from MSDEV and some from Eclipse
- Codelite supports Subversion, MinGW and its setup is very fast and
straightforward
- Codelite has many positive reviews
That's why discussion in topic "migration to Eclipse" continued in
"migration to Codelite"
Phoenix
it....moreover in the human workspace there is duplication of code.
The include files for generic should be included from generic project
rather than copying the same files in aiengine/src/include.
For sockets try also including -lwsock32 library. You can copy that
static lib from MINGW installation folder under /lib folder.
Phoenix
to codelite". Thanks
Habbit
generic include and binaries are intentionally separated from
aiengine.
Rare updates are expected in generic - let's consider it as 3d-party
library.
Currently it is mature enough (though cannot say good) and I think it
is better to have boundary to not allow immaturity to expand.
I don't think it is good to have direct inter-project links.
Imagine you have relatively big components A and B, A depends from B
(in practice we usually have C which also depends from B).
Also you have mainline (HEAD) versions of components and release
(stable, verified) versions.
So in theory you can have Arel x Brel, Ahead x Brel, Ahead x Bhead or
Arel x Bhead. In terms of stability, Ahead x Bhead should be avoided,
because unstable changes in Bhead (maybe done for C or for some future
change of A) multiplied by unstable changes in Ahead will create extra
instability and inconvenience for other team members.
Currently we have Ahead x Brel configuration as std development
framework.
In local workspace (w/o commit) you can have Ahead x Bhead if you have
correlated changes.
Re sockets - in aiengine -lws2_32 is working fine.
What is the difference?
-lwsock32 - is it winsock 1.0?
Both projects are using winsock2.h header.
Habbit
compare:
1.
----------Building project:[ aiconsole - Debug ]----------
g++ -o ./bin/aiconsole-debug.exe ./build/debug/main.o "-L." "-LK:/
projects/ai2svn/human/aiconsole/lib/debug" -laiapi.debug -lws2_32
----------Build Ended----------
0 errors, 0 warnings
2.
----------Building project:[ aiconsole - Debug ]----------
g++ -o ./bin/aiconsole-debug.exe ./build/debug/main.o "-L." "-LK:/
projects/ai2svn/human/aiconsole/lib/debug" -lws2_32 -laiapi.debug
K:/projects/ai2svn/human/aiconsole/lib/debug/libaiapi.debug.a
(aiapisocket.o):K:/projects/ai2svn/human/aiapi/src/aiapisocket.cpp:21:
undefined reference to `WSAStartup@8'
The link order has impact on results!
This is quite unexpected for me...
Habbit
... and deleted migration fodlers.
tried server run - it is ok now - behaves as before.
still not tried debug...
Phoenix
the IDE now and the project configuration. Yes i agree that we should
treat generic as well as aiapi as 3rd party and keep them separate in
the code base.
about the sequence of library ..its interesting and a learning for
me...creating a separate thrad for all the learnings.....
will copy yours in the same.
Habbit
I have 2 problems:
For some reason aiengine incremental build always performs link stage.
When starting debugger, it reports that unable set breakpoint.
If turn on full debug logging, you can see:
DEBUG>>00000012-break-insert "\"K:/projects/ai2svn/human/aiengine/src/
main.cpp:10\""
DEBUG>>00000012^error,msg="No source file named K:/projects/ai2svn/
human/aiengine/src/main.cpp."
Breakpoint creation unsuccessful
ERROR: failed to place breakpoint: "^error,msg="No source file named
K:/projects/ai2svn/human/aiengine/src/main.cpp.""
No source file named K:/projects/ai2svn/human/aiengine/src/main.cpp.
So debugger adds extra "." to the end of the file.
It is open question why...