for debug configuration or obj-release for release configuration.
- instead of gui,iodev,cpu,etc. having their own obj-debug directory, it
all goes into a single output directory for the whole project. I did
this when I was having trouble getting any debug information from the
static library directories to be recognized in the VC++ debugger. Maybe
it's not necessary.
- changed the compile arguments for the debug config so that debug symbols
are created. We used to have:
debug config: /G6 /MT /W4 /GX /O2
Now it says this:
debug config: /MTd /W3 /GX /Z7 /Od
I guess I haven't tested the release config yet, but the debug config is
working great now. I even have symbolic debugging of functions in the
gui/iodev/cpu libraries.
removed! I used this trick sometimes to check that a function returned
what I expected, like assert (func () == 0), but this caused the
func() to never get called. Oops.
device and disk file for a while. Even though its version of
read_toc is minimal, in fact I would say broken, it lets people use
an ISO disk file as a cdrom.
- in this revision, I wrote the "unix equivalent" of the win32 code, including
the broken version of read_toc. Now win32 and unix should act very similar
when they encounter an ISO disk image.
- one important improvement is in read_toc, I have added "*length=1" for both
win32 and unix, since otherwise the function returns random junk for the
length of the TOC. I also tried "*length=0" and that created the "lost
interrupt" behavior that psyon has been trying to get rid of...I changed it
back to *length=1 of course and left a note to him in that bug report.
will dump core, and the core can sometimes be loaded into the debugger
to give some hints as to what is going on. This could be especially
useful when the bochs debugger is off.
- try to do something reasonable when abort() doesn't exist on the system.
which were generated with gcc -MM to the end of each Makefile.in
so that make understands which files depend on which. Basically,
everything depends on bochs.h, which depends on everything, which
is not ideal.
- rework the order of initialization with and without the control panel.
The thing that was bothering me most was the command line options were
being processed after the user had set everything in the control panel.
This is clearly not what's expected--the command line options should
affect the startup defaults of the control panel, but whatever the user
chooses in the cpanel menus is the final choice.
- if the control panel (config interface) is not wanted, the user can
put "-nocp" or "-nocontrolpanel" as the FIRST argument on the command
line. Also, the "-psn" option which is automatically passed in by
MacOS X when you doubleclick the application causes the control panel
to be disabled. In this case, the order of operations is:
1. read bochsrc
2. parse command line options.
- if the control panel is enabled (default), the command line options are
parsed to provide the startup defaults for the control panel, but the
control panel settings are the final answer. So the order is:
1. parse command line options
2. run control panel (if user chooses, he can read bochsrc from menus)
- I haven't tested command line options with the debugger yet.
of just panicing. In particular, if the logical sector is out of bounds
or the disk image cannot be read/written at the desired offset, we now
abort the ATA command and return an error code. Many of the old BX_PANIC
messages are turned to BX_ERROR, so they will still appear in the
log, but now the device model will try to communicate this fact to
the OS instead of simply giving up.
that TICK is called so I put a trace call just before each TICK.
This seems best, since the trace has a chance to print before the tick
can trigger time-based events elsewhere in the system.
> Attached is a patch to get Bochs 1.2.1 compiling with the Sun Workshop 6
> compiler suite...
>
> Plain - out the box - the build fails with:
> -- snip --
> "pic.cc", line 503: Error: Badly formed constant expression.
> "pic.cc", line 533: Error: Badly formed constant expression.
> -- snip --
> It looks that the preprocessor expects an integer in these cases and not
> a string...
>
>
> Additionally I added some fixes to get rid of the Zillion compiler
> warnings like "Warning: String literal converted to char* in
