Mail Archives: djgpp-workers/2001/01/02/11:03:20
I've been running go32 and go32-v2 in every combination according to
src/stub/go32-v2.c:
/* If you want to change this, remember to test it with all the
various ways `go32' can be called. The following is the
list of different cases I know about:
x go32 is called from DOS prompt without any arguments
x go32 is called from DOS prompt to run unstubbed COFF image of a v2 program
x(1) go32 is called from DOS prompt to debug a v1 COFF image (go32 -d)
x v2's Make calls go32 to run unstubbed COFF image of a v2 program
x v1's Make calls go32 to run unstubbed COFF image of a v2 program
x 16-bit Make calls go32 to run unstubbed COFF image of a v2 program
x v2's Make calls go32 to run unstubbed COFF image of a v1 program
x v1's Make calls go32 to run unstubbed COFF image of a v1 program
x 16-bit Make calls go32 to run unstubbed COFF image of a v1 program
x v1 .EXE program is called from the DOS prompt
x v1 symlink is called from the DOS prompt
x v2's Make calls a v1 .EXE program
x v1's Make calls a v1 .EXE program
x 16-bit Make calls a v1 .EXE program
x(2) when v2's Make calls go32, you need to test command lines which
are both shorter and longer than the DOS 126-character limit
Note that there is nothing special about Make, it just serves as an
example of one program that calls another. It is convenient to use
Make to test the above cases, because Make is available as both v2
and v1 executable and as a 16-bit program, and because it can be
easily used to call different programs with different command lines
by tweaking a Makefile. */
(1) go32-v2 can't find edebug32 or ed32-dpm unless it's in the
same directory or the full path to it is given.
(2) Works exactly like djdev203.zip version, i. e. almost all
cases except the reported one (make bug?).
Have I understood the setup correctly:
1. A v2 setup.
2. Put go32.exe in a directory in the path, but after DJGPP/BIN.
3. Rename go32-v2.exe to go32.exe.
4. Borlands real-mode make (called MAKER.EXE) is a 16-bit make.
If everyone is happy with this, I'll commit the following patch,
which fixes that reserved field in the header and changes dates to
YYYY-MM-DD format.
Right,
MartinS
Index: src/stub/djasm.y
===================================================================
RCS file: /cvs/djgpp/djgpp/src/stub/djasm.y,v
retrieving revision 1.5
diff -p -u -r1.5 djasm.y
--- djasm.y 1999/06/03 17:27:41 1.5
+++ djasm.y 2001/01/02 15:29:28
@@ -1362,11 +1362,20 @@ int main(int argc, char **argv)
exe[22] = 0; /* relative CS */
exe[23] = 0;
+ /* These must be zero, otherwise they are interpreted as an offset to
+ a "new executable" header. */
+ exe[60] = 0;
+ exe[61] = 0;
+ exe[62] = 0;
+ exe[63] = 0;
+#define INFO_TEXT_START (64)
+
time(&now);
- sprintf(exe+28, "\r\n%s generated from %s by djasm, on %.24s\r\n", argv[2], argv[1], ctime(&now));
+
+ sprintf(exe+INFO_TEXT_START, "\r\n%s generated from %s by djasm, on %.24s\r\n", argv[2], argv[1], ctime(&now));
if (copyright)
- strncat(exe+36, copyright, 476-strlen(exe+36));
- strcat(exe+36, "\r\n\032");
+ strncat(exe+INFO_TEXT_START, copyright, (512-3-INFO_TEXT_START)-strlen(exe+INFO_TEXT_START)); /* -3 for the following line: */
+ strcat(exe+INFO_TEXT_START, "\r\n\032");
if (argv[2] == 0)
{
@@ -2510,20 +2519,20 @@ void add_rcs_ident(char *buf)
struct tm *tm;
time(&now);
tm = localtime(&now);
- sprintf(tmp, "%cId: %s built %02d/%02d/%02d %02d:%02d:%02d by djasm $\n",
+ sprintf(tmp, "%cId: %s built %04d-%02d-%02d %02d:%02d:%02d by djasm $\n",
'$', inname,
+ tm->tm_year + 1900,
tm->tm_mon + 1,
tm->tm_mday,
- tm->tm_year,
tm->tm_hour,
tm->tm_min,
tm->tm_sec);
add_copyright(tmp);
- sprintf(tmp, "@(#) %s built %02d/%02d/%02d %02d:%02d:%02d by djasm\n",
+ sprintf(tmp, "@(#) %s built %04d-%02d-%02d %02d:%02d:%02d by djasm\n",
inname,
+ tm->tm_year + 1900,
tm->tm_mon + 1,
tm->tm_mday,
- tm->tm_year,
tm->tm_hour,
tm->tm_min,
tm->tm_sec);
- Raw text -