Hi,
When I was trying to use md5check.exe in R on Windows, I found that it is
actually crashed for R-2.7.2 and R-2.6.2 (R-2.5.1 seems to be ok).
As I looked into the code (src/gnuwin32/front-ends/md5check.c), I think that
there might be several bugs in the code:
(1) A "segmentation fault" occurred in
read_unist_file("unins000.dat"). I
guess that the file pointer didn't stop when it reached EOF.
(2) For some reason, the lastest version of "snprintf" appended
'\0' to the
string, so the following line:
for(j = 0; j < 16; j++) snprintf (out+2*j, 2, "%02x",
resblock[j]);
produced a (MD5) strings with '\0' at every even position.
(3) When there is no unins000.dat file, md5check.exe won't report any
missing files.
I think the last two "bugs" can be fixed by the following minor
changes:
============================
--- R-2.7.2/src/gnuwin32/front-ends/md5check.c (revision 167)
+++ R-2.7.2/src/gnuwin32/front-ends/md5check.c (working copy)
@@ -157,7 +157,7 @@
#ifdef DEBUG
printf("missing\n");
#endif
- if(found) {
+ if(found||nnames==0) {
fprintf(stderr, "file %s: missing\n", line+34);
miss++;
}
@@ -171,7 +171,7 @@
#endif
continue;
} else {
- for(j = 0; j < 16; j++) snprintf (out+2*j, 2, "%02x",
resblock[j]);
+ for(j = 0; j < 16; j++) snprintf (out+2*j, 3, "%02x",
resblock[j]);
out[32] = '\0';
if(strcmp(onfile, out) == 0) {
#ifdef DEBUG
============================
I didn't fix the first one, because I am not familiar with the format of
unins000.dat.
Thanks,
Alex
###########################
Alex Chen, Ph.D.
REvolution Computing
1100 Dexter Ave. N, Suite 200
Tel: 206-577-4778
###########################
[[alternative HTML version deleted]]
