"Bill Page" <[EMAIL PROTECTED]> writes:
[...]
| I think the build is ok now and I can start AXIOMsys. But I
| do get an error when I try hyperdoc:
|
| -bash-3.00$ axiom
| (HyperDoc) read_ht_db: No ht.db file found
OK, I found where the bug comes from. It is a very stupid bug.
A while ago, I added support for relative paths -- that was a
necessary step to make axiom.build-improvements buildable on Windows.
I believe that is when HyperDoc stopped working. The reason?
Well, a very sloppy logic in the HyperDoc source code.
More precisely, the function src/hyper/addfile.c:build_ht_filename
static int
build_ht_filename(char *fname, char *aname, char *name)
{
char cdir[256];
char *c_dir;
char *HTPATH;
char *trace;
char *trace2;
int ht_file;
if (cwd(name)) {
/* user wants to use the current working directory */
c_dir = (char *) getcwd(cdir, 254);
strcpy(fname, c_dir);
/* Now add the rest of the filename */
strcat(fname, "/");
strcat(fname, &name[2]);
/** now copy the actual file name to addname **/
for (trace = &name[strlen(name)]; trace != name &&
(*trace != '/'); trace--);
if (trace == name) {
fprintf(stderr, "ht_open_file: Didn't expect a filename like %s\n",
name);
exit(-1);
}
trace++;
strcpy(aname, trace);
/** add the .ht extension if needed **/
extend_ht(aname);
extend_ht(fname);
/* Now just try to access the file */
return (access(fname, R_OK));
}
else if (pathname(name)) {
/* filename already has the path specified */
strcpy(fname, name);
/** now copy the actual file name to addname **/
for (trace = &name[strlen(name)]; trace != name &&
(*trace != '/'); trace--);
if (trace == name) {
fprintf(stderr, "ht_open_file: Didn't expect a filename like %s\n",
name);
exit(-1);
}
trace++;
strcpy(aname, trace);
/** add the .ht extension if needed **/
extend_ht(aname);
extend_ht(fname);
/* Now just try to access the file */
return (access(fname, R_OK));
}
else {/** If not I am going to have to append path names to it **/
HTPATH = (char *) getenv("HTPATH");
if (HTPATH == NULL) {
/** The user does not have a HTPATH, so I will use the the directory
$AXIOM/share/hypertex/pages/ as the default path ***/
char *spad = (char *) getenv("AXIOM");
if (spad == NULL) {
fprintf(stderr,
"ht_file_open:Cannot find ht data base: setenv HTPATH or AXIOM\n");
exit(-1);
}
HTPATH = (char *) halloc(1024 * sizeof(char), "HTPATH");
strcpy(HTPATH, spad);
strcat(HTPATH, "/share/hypertex/pages");
}
/** Now that I have filled HTPATH, I should try to open a file by the
given name **/
strcpy(aname, name);
extend_ht(aname);
for (ht_file = -1, trace2 = HTPATH;
ht_file == -1 && *trace2 != '\0';) {
for (trace = fname; *trace2 != '\0' && (*trace2 != ':');)
*trace++ = *trace2++;
*trace++ = '/';
*trace = 0;
if (!strcmp(fname, "./")) {
/** The person wishes me to check the current directory too **/
getcwd(fname, 256);
strcat(fname, "/");
}
if (*trace2)
trace2++;
strcat(fname, aname);
ht_file = access(fname, R_OK);
}
return (ht_file);
}
}
The comment fragment
if (cwd(name)) {
/* user wants to use the current working directory */
is deceptive because here is how it is deduced that user wanst the
current working directory
#define cwd(n) ((n[0] == '.' && n[1] == '/')?(1):(0))
That is only the first two characters are tested!
It happens that in the build machinery we use things like
./../../target/i686-ps-linux-gnu
and teh cwd() macro thinks that that is current working directory!
Ahem.
The other fragment
else if (pathname(name)) {
/* filename already has the path specified */
is equally deceptive.
-- Gaby
_______________________________________________
Axiom-developer mailing list
Axiom-developer@nongnu.org
http://lists.nongnu.org/mailman/listinfo/axiom-developer
