I got page faults while compiling C++ files in Borland BC4.52
and Microsoft VC1.52 on the latest CVS build. I have tested
the same files on DosBox 0.63 release and found no this issue.
I hope the next release will fix it.
indeed interresting.
did you try it with core_full as well ?
Water flows down the stream
How to ask questions the smart way!
wrote:indeed interresting.
did you try it with core_full as well ?
I have tested all the three cores. All give the same result.
Can I use paging.cpp of 0.63 release in CVS?
It's worth a try. It may be useful to do a diff and see if anything sticks out.
Also could you post some debug logs?
wrote:Also could you post some debug logs?
Here is the image of debug session.
Does it behave if you set the memory to some large values
like 64 or so? And did exchanging the paging.cpp-file
(maybe cpu.cpp as well) work?
If you have a debugger-enabled compile (that is the dosbox
debugger) you can generate debug logs that contain some
information about the pagefaults (have a look at paging.cpp
to have it output even more infos like register values).
wd
wrote:If you have a debugger-enabled compile (that is the dosbox debugger) you can generate debug logs that contain some information a […]
If you have a debugger-enabled compile (that is the dosbox
debugger) you can generate debug logs that contain some
information about the pagefaults (have a look at paging.cpp
to have it output even more infos like register values).
wd
To keep the exe file small, I build DOSBOX with required options only.
Even so, the release target produced by VS2005 is almost 900K. As I can debug it within IDE, I don't want to build it with debugger-enabled.
wrote:Does it behave if you set the memory to some large values
like 64 or so? And did exchanging the paging.cpp-file
(maybe cpu.cpp as well) work?
No. I did the tests on default settings come with CVS builds.
wrote:It's worth a try. It may be useful to do a diff and see if anything sticks out.
wrote:And did exchanging the paging.cpp-file
(maybe cpu.cpp as well) work?
Yes. You are right. I replaced paging.cpp with the one from 0.63 release and did a CVS build. It works.
> To keep the exe file small, I build DOSBOX with required options only.
> Even so, the release target produced by VS2005 is almost 900K. As I can
> debug it within IDE, I don't want to build it with debugger-enabled.
The VS-Debugger is no replacement for the dosbox debugger.
The latter is capable of debugging the dos-applications (in
your case BC4/VC1.5) and supplies additional information
like unemulated aspects.
If you can provide either debug logs of the dosbox debugger
running the program, or examples (like the executables needed
to reproduce the problem, but be aware of copyright issues)
it might be possible to fix it.
wd
cyberwalker: So the next thing I'd try is to see what changed in that file between 0.63 and CVS that might cause the behavior you're seeing.
wrote:cyberwalker: So the next thing I'd try is to see what changed in that file between 0.63 and CVS that might cause the behavior you're seeing.
I think you are right, but I don't understand paging of DOSBOX.
I got bc5 and that one still works
maybe you can attach a sample sourcefile ?
as I did a mere hello world
Water flows down the stream
How to ask questions the smart way!
wrote:I got bc5 and that one still works
maybe you can attach a sample sourcefile ?
as I did a mere hello world
The test source requires Borland Power Pack, it can not be built in BC5.x. I have got the log files by paging.cpp of CVS and 0.63 release(Thank wd). The log files are attached to the original post.
Rumours have it that wd fixed the problem. It should be in the cvs source.
Water flows down the stream
How to ask questions the smart way!
wrote:Rumours have it that wd fixed the problem. It should be in the cvs source.
Yes. I am testing this fix. No issues found as yet.
