CloverBootloader/BaseTools/Source/C/VfrCompile/Pccts/NOTES.msvc
2019-09-03 12:58:42 +03:00

190 lines
7.1 KiB
Plaintext

Microsoft Visual C Stuff
[Tom Moog 2-Oct-98
Users of Microsoft Visual C++ should download a separate
ready-to-run zip file from my web site. It contains
binaries, static library, and a sample project.
]
[
Two notes added by Tom Moog 23-Sep-97. I believe the *.dsp and
*.mak files that were once at the end of this file are now obsolete.
The following MSVC .dsp and .mak files for pccts and sorcerer
were contributed by Stanislaw Bochnak (S.Bochnak@microtool.com.pl)
and Jeff Vincent (jvincent@novell.com)
PCCTS Distribution Kit
----------------------
pccts/antlr/AntlrMSVC50.dsp
pccts/antlr/AntlrMSVC50.mak
pccts/dlg/DlgMSVC50.dsp
pccts/dlg/DlgMSVC50.mak
pccts/support/genmk/watgenmk.mak
pccts/support/msvc.dsp
Sorcerer Distribution Kit
-------------------------
pccts/sorcerer/SorcererMSVC50.dsp
pccts/sorcerer/SorcererMSVC50.mak
pccts/sorcerer/lib/msvc.dsp
I do not have an MS based computer. If you discover problems
please report them so as to save trouble for others in the future.
]
[
Modified by Terence Parr (September 1995) to change .C to .cpp
]
[
This file contains notes on MSVC for Windows NT console execs by Dave
Seidel and an explanation of flags etc.. by John Hall; good luck,
Terence
]
===============================================================================
Date: Sat, 31 Dec 1994 11:40:36 -0500 (EST)
From: David Seidel <75342.2034@compuserve.com>
I've succesfully build 1.31b3 with djgpp for DOS and MSVC 2.0 for Windows
NT. The only (minor) problem I had was that GNU make (version 3.71, in the
djgpp port) complained about "multiple targets" in both the antlr and dlg
makefiles. I got around the error by, in each makefile, commenting out the
$(SRC) dependency, for example:
antlr: $(OBJ) #$(SRC)
I don't know why this is happenning, since you haven't changed that part of
the makefile at all, and I think this used to work ok...
Here are the makefiles I built from within the MSVC 2.0 environment for antlr
and dlg and Windows NT console executables. Please feel free to pass them
on. Of course, as soon as 1.31 "goes gold", I will send you nice new
binaries. I'm not going to bother to keep doing both Borland and djgpp for
DOS however. Instead, I'll just keep the djgpp version up to date and also
provide WinNT binaries.
Dave
===============================================================================
How to port PCCTS 1.10 (and 1.32 hopefully) to Visual C++
By
John Hall <jhall@ivy.wpi.edu>
Here is how to compile an ANTLR grammar in Visual C++. These steps
describe how to have your ANTLR grammar parse the input file the user
selects when they choose File Open in your Windows application. (Even
if you aren't using Visual C++, the steps should be portable enough to
other compilers.)
* Make sure that ANTLR and DLG generate ANSI code (use the -ga
switch).
* Set the following compiler flags in Visual C++ (these are in the
Memory Model category of the compiler options in the Project
Options menu):
FLAG MEANING
==== ==============================================================
/AL Large memory model (multiple data segments; data items must be
smaller than 64K).
/Gtn Allocates all items whose size is greater than or equal to n
in a new data segment. (I let n be 256: /Gt256.)
/Gx- All references to data items are done with far addressing in
case they are placed in a far segment.
* Add the following member variable to the attributes section of your
derived CDocument class (you will need to make sure you also
include stdio.h):
FILE *fp;
* Add the following method to your derived CDocument class:
BOOL CAppDoc::OnOpenDocument(const char* pszPathName)
{
// Call CDocument's OnOpenDocument to do housekeeping for us
// DON'T add anything to the loading section of Serialize
if (!CDocument::OnOpenDocument(pszPathName))
return FALSE;
// Open input file
if ((fp = fopen(pszPathName, "r")) == NULL)
return FALSE;
// Parse input file
ANTLR(start(), fp);
// Close input file
fclose(fp);
return TRUE;
}
(Note: additional code may be necessary, depending on your parser.
For example, if your parser uses PCCTS's symbol table library, you
will need to insert calls to zzs_init and zzs_done.)
* Compile the generated C files as C++ files. (I renamed the files
to have a .CPP extension to fool Visual C++ into thinking they were
C++ files. One might also use the /Tp switch, but that switch
requires you separately include the filename.) [I used this step
as an easy out for all the external linking errors I was getting
that I couldn't fix by declaring things extern "C".]
* Make sure the __STDC__ portion of the generated files gets
compiled. (Either define __STDC__ yourself or else change all
occurrences of __STDC__ to __cplusplus in the generated files. You
can define __STDC__ in the Preprocessor category of the compiler
options.)
================================================================
= Note 23-Sep-97: This is probably not necessary any more. =
= With 1.33MRxxx the use of __STDC__ was replaced with the =
= macro __USE_PROTOS to control the compilation of prototypes. =
================================================================
That last step is important for Visual C++, but may not apply to other
compilers. For C++ compilers, whether __STDC__ is defined is
implementation dependent (ARM, page 379). Apparently, Visual C++ does
not to define it; it also does not support "old style" C function
definitions (which is okay, according to page 404 of the ARM). Those
two things together caused problems when trying to port the code.
When it saw this:
#ifdef __STDC__
void
globals(AST **_root)
#else
globals(_root)
AST **_root;
#endif
it skipped the __STDC__ section and tried to process the "old style"
function definition, where it choked.
When you finally get your parser to compile and link without error,
you may get General Protection Fault errors at run time. The problem
I had was that a NULL was passed to a variable argument function
without an explicit cast. The function grabbed a pointer (32-bits)
off the stack using va_arg, but the NULL was passed silently as the
integer 0 (16 bits), making the resulting pointer was invalid. (This
was in PCCTS's sample C parser.)
There is one other thing I might suggest to help you avoid a run-time
error. Make sure you redefine the default error reporting function,
zzsyn. To do this, put "#define USER_ZZSYN" in your #header section
and put your own zzsyn somewhere. You can then pop up a MessageBox or
print the error to some output window.
===============================================================================