The Watcom C/C++ libraries are located under the /usr/lib
directory.
Due to the many code generation strategies possible in the 80x86
family of processors, a number of versions of the libraries are
provided. You must use the libraries that coincide with the particular
code generation strategy or model that you've selected.
For the type of code generation strategy or model that you intend to
use, refer to the description of the
m{f,s,m,c,l,h}
memory model compiler option.
The various code models supported by Watcom C/C++ are described in the
chapters 16-bit Memory
Models (in the 16-bit Topics supplement), and
32-bit Memory Models (in this manual).
We've selected a simple naming convention for the libraries that are
provided with Watcom C/C++.
Letters are affixed to the file name to indicate the particular
strategy with which the modules in the library have been compiled.
The following letters apply only to 16-bit libraries, and indicate the
memory model used:
| Letter
| Memory Model
| Library
|
|---|
| s
| small (small code, small data)
| clibs.lib
|
| m
| medium (big code, small data)
| clibm.lib
|
| c
| compact (small code, big data)
| clibc.lib
|
| l
| large (big code, big data)
| clibl.lib
|
| h
| huge (big code, huge data)
| clibh.lib
|
The following letters apply to 32-bit libraries (for the flat and small memory
models), and indicate the compiler option used:
| Letter
| Option
| Library
|
|---|
| 3r
| 3r, 4r or 5r
| clib3r.lib
|
| 3s
| 3s, 4s or 5s
| clib3s.lib |
A portion of the 16-bit Watcom C Library is also stored in a
memory-resident library called the system shared library.
On multi-tasking systems, it makes sense that commonly-used library
routines such as read() and write() be shared among
processes. By sharing the same code, the memory requirement for applications is
reduced. The functions in the shared library are independent of the memory
model, so they can be used by any small/large code, small/large/huge data
applications.
The Watcom C++ class library routines are described in the
Watcom C++ Class Library Reference.
The Watcom C++ 16-bit Class Libraries for various combinations of compiler
options and memory models are listed below:
| Classes
| Option
| Model
| Library
|
|---|
| iostream, string
| | small
| plibs.lib
|
| | | medium
| plibm.lib
|
| | | compact
| plibc.lib
|
| | | large
| plibl.lib
|
| | | huge
| plibh.lib
|
| Complex
| fpc
| small
| cplxs.lib
|
| | | medium
| cplxm.lib
|
| | | compact
| cplxc.lib
|
| | | large
| cplxl.lib
|
| | | huge
| cplxh.lib
|
| Complex
| fpi or fpi87, 7
| small
| cplx7s.lib
|
| | | medium
| cplx7m.lib
|
| | | compact
| cplx7c.lib
|
| | | large
| cplx7l.lib
|
| | | huge
| cplx7h.lib |
These libraries are independent of the operating system.
The ``7'' designates a library compiled with the 7 option.
The Watcom C++32-bit Class Libraries for various compiler
options for the small and flat memory models are listed below.
| Classes
| Compiler Option
| Library
|
|---|
| iostream, string
| 3r, 4r or 5r
| plib3r.lib
|
| | 3s, 4s or 5s
| plib3s.lib
|
| Complex
| fpc, 3r, 4r or 5r
| cplx3r.lib
|
| | fpc, 3s, 4s or 5s
| cplx3s.lib
|
| Complex
| 7, fpi or fpi87, and 3r,
4r or 5r
| cplx73r.lib
|
| | 7, fpi or fpi87, and 3s,
4s or 5s
| cplx73s.lib |
These libraries are independent of the operating system.
The ``3r'' and ``3s'' suffixes refer to the
argument-passing convention used.
The ``7'' designates a library compiled with the 7 option.
In general, a Math library is required when floating-point
computations are included in the application.
The Math libraries are operating-system independent.
The Math libraries are placed under the /usr/lib directory.
The following situations indicate that one of the Math libraries
should be included when linking the application.
- If an application uses one or more of the functions described in the
math.h header file, then a Math library must be included.
-
If an application is linked and the message:
_fltused_ is an undefined reference
appears, then a Math library must be included.
- (16-bit only) If an application is linked and the message:
__init_87_emulator is an undefined reference
appears, then one of the modules in the application was compiled with
one of the fpi, fpi87, fp2, fp3
or fp5 options.
If the fpi option was used, the 80x87 emulator library
(emu87.lib) or the 80x87 fixup library
(noemu87.lib) should be included when linking the
application.
If the fpi87 option was used, the 80x87 fixup library
noemu87.lib should be included when linking the application.
- (32-bit only) If an application is linked and the message:
__init_387_emulator is an undefined reference
appears, then one of the modules in the application was compiled with
one of the fpi, fpi87, fp2,
fp3 or fp5 options.
If the fpi option was used, the 80x87 emulator library
(emu387.lib) should be included when linking the
application.
If the fpi87 option was used, the empty 80x87 emulator library
noemu387.lib should be included when linking the
application.
Normally, the compiler and linker will automatically take care of this.
Simply ensure that the WATCOM
environment variable includes the location of the Watcom C/C++ libraries.
One of the following Math libraries must be used if any of the modules
of your application has been compiled with one of the Watcom C/C++ compiler
fpi, fpi87, fp2, fp3 or
fp5 options, and your application requires
floating-point support for the reasons given above.
The 16-bit libraries are as follows:
| Library
| Memory Model
|
|---|
| math87s.lib
| small
|
| math87m.lib
| medium
|
| math87c.lib
| compact
|
| math87l.lib
| large
|
| math87h.lib
| huge
|
| noemu87.lib
| any
|
| emu87.lib
| any |
The 32-bit libraries for the small and flat memory models are as follows:
| Library
| Compiler Option
|
|---|
| math387r.lib
| 3r, 4r or 5r
|
| math387s.lib
| 3s, 4s or 5s
|
| emu387.lib
| |
The fpi compiler option causes an 80x87 numeric data processor
emulator to be linked into your application, in addition to any 80x87 math
routines that were referenced.
For QNX, there is a common 80x87 emulator task that's used so that
there's only one copy of the emulator in memory at any one time.
This emulator decodes and emulates 80x87 instructions when an 80x87
isn't present in the
system.
When the fpi87 compiler option is used exclusively, the emulator
isn't included.
In this case, the application must be run on personal computer systems
equipped with the numeric data processor.
One of the following Math libraries must be used if any of the modules
of your application has been compiled with the fpc option of the
Watcom C/C++ compiler, and your application requires floating-point support
for the reasons given above.
The following Math libraries include support for floating-point, which
is done out-of-line through runtime calls.
The 16-bit libraries are as follows:
| Library
| Memory Model
|
|---|
| maths.lib
| small model
|
| mathm.lib
| medium model
|
| mathc.lib
| compact model
|
| mathl.lib
| large model
|
| mathh.lib
| huge model |
The 32-bit libraries are as follows:
| Library
| Compiler Option
|
|---|
| math3r.lib
| 3r, 4r or 5r
|
| math3s.lib
| 3s, 4s or 5s |
Applications that are linked with one of these libraries don't
require a numeric data processor for floating-point operations.
If one's present in the system, it's used; otherwise
floating-point operations are simulated in software.
Source files are included in the package for the Watcom C/C++ application
startup (or initialization) sequence.
For QNX, these files are located in the directory
/usr/lib/src/startup.
The following is a summary list of the startup files for QNX:
- cmain.c
- final part of initialization sequence
- mdef.inc
- macros included by cstart.asm
- models.inc
- included by cstart_*.asm
- start.asm
- included by cstart_*.asm
- start_c.asm
- startup for compact memory model
- start_h.asm
- startup for huge memory model
- start_l.asm
- startup for large memory model
- start_m.asm
- startup for medium memory model
- start_s.asm
- startup for small memory model
The assembler file cstart.asm contains the first part of the
initialization code, and the remainder is continued in the file
cmain.c.
The assembler files, cstart_*.asm, define the type of memory
model, and include cstart.asm. It's cmain.c
that calls your mainline routine, main().
For QNX, these files are located in the directory
/usr/lib/src/startup.
The following is a summary list of the startup files for QNX:
- cstrt386.asm
- startup for small memory model
- mdef.inc
- macros included by cstrt386.asm
- cmain.c
- final part of initialization sequence
The assembler file cstrt386.asm
contains the first part of the initialization code, and the remainder
is continued in the file cmain.c. It's
cmain.c that calls your mainline routine main().
The source code is provided for those who wish to customize the
initialization sequence for special applications.