When running, CodeView presents the user with several
windows that can be tiled, moved, sized and otherwise manipulated via the keyboard or mouse, with CodeView 4.x providing a richer interface. Some of the windows include: • Code window – the code window showed the currently debugged code in its
source code context. • Data window – a
hexadecimal dump of a user-specified memory area. • Watch window – a contextual display of variables by name. • Locals window – a contextual display of variables local to the current function. • Command window – user commands (using the same or similar syntax as DEBUG and SYMDEB) could be entered here. • Assembly window – the assembly (machine code) was displayed, allowing for single-stepping through functions. • Register window – to visualize the
80x86 register contents, including segments, flags and the
FPU (CodeView existed before
MMX and other
SIMD extensions). • Output window – a window showing startup operations and debugging information relating to breakpoints, hardware breaks (interrupt 0 and 3), etc.
Features • 386 mode – 8086, 80286 and 80386 and later processors. 386 enhanced mode is activated by a menu option, allowing for 32-bit registers and disassembly. • Monochrome monitor support – allows debugging on either a single color (CGA, EGA or VGA) monitor with page/memory swapping between the user application and the CodeView screen, or using a separate monochrome monitor. The monochrome monitor exists in memory address space 0xb0000, while the color monitor exists at 0xb8000 for text and 0xa0000 for graphics. Use of the monochrome monitor with its separate memory address space allows debugging graphics applications without affecting the display, as well as all text modes. Monochrome monitors are limited to 25 lines, whereas color monitors allow 25, 43 or 50 line mode, allowing for more information on the screen at the same time. Creating symbolic debugging output, which allows memory locations to be viewed by their programmer-assigned name, along with a program database showing the source code line related to every computer instruction in the binary executable, is enabled by the command line switch -Zi given to the compiler, and -CO given to the linker. Variants like -Zs and -Zd provide lesser information, and smaller output files which, during the early 1990s, were important due to limited machine resources, such as memory and hard disk capacity. Many systems in those days had 8MB of memory or less. CodeView handles all program models, including TINY, SMALL, COMPACT, MEDIUM, LARGE and HUGE, with TINY (DOS-based .COM files) having their symbolic debugger information stored in a separate file, with all of the other .EXE formats containing the symbolic information directly inside the executable. This often introduced a notable size increase, and it therefore became desirable for some developers to use #pragma switches within their C (and later C++) source code to prevent the majority of the application from having symbolic output, and instead limiting that output to only those portions which required it for current debugging. CodeView version 3.x and 4.x introduced various transport layers, which removed some of the memory space limitations to this form of symbolic debugging. Typically the debugger runs in the lower 640KB memory space alongside the application being debugged, which greatly decreases the amount of memory available to the application being debugged. The transport layer allows only a stub to exist in main memory, while the bulk of the debugger code resides in EMS or XMS (memory above the 1 MB barrier, or outside of the normal 0 KB - 640 KB address space typically used by DOS programs). CodeView also came with a CVPACK command-line utility, which can reduce the size of the CodeView-generated information internally, while still retaining full symbolic access to data. ==Visual C++ support==