[tlbuild] dvi test?

[John Hawkinson]

John Hawkinson <jhawk at alum.mit.edu> wrote on Sun, 30 Jan 2022
at 22:20:31 EST in <YfdVfxRWBAurmv2W at louder-room.local>:

>      if (fstat(fileno(m_dvi_fp), &fstatbuf) != 0 || S_ISDIR(fstatbuf.st_mode)) {	/* if it's a directory */

So, this code seems wrong.
fstat(2) returns 0 on success, and -1 on failure. Normally we test its return value against -1.

So here, if fstat(2) succeeds, then first half of the condition (!= 0) is false, and then we try to evaluate S_ISDIR() to see if its true.
But it's not reasonable to call S_ISDIR on fstatbuf, since fstatbuf is only written to if fstat(2) had succeeded.

If fstat(2) fails, then the if condition shortcircuits (because -1 != 0), and then we set *errflag = FILE_IS_DIRECTORY, but
that's wrong because we don't know why fstat(2) failed and whether S_ISDIR() would have returned true.

That said, I don't think this is really our problem. But I applied this patch:

--- a/texk/xdvik/dvi-init.c
+++ b/texk/xdvik/dvi-init.c
@@ -1242,7 +1242,11 @@ file_exists_p(const char *path, dviErrFlagT *errflag)
     /*      fprintf(stderr, "after internal_open_dvi2: xfopen\n"); */
     
     /* shouldn't happen */
-    if (fstat(fileno(m_dvi_fp), &fstatbuf) != 0 || S_ISDIR(fstatbuf.st_mode)) {	/* if it's a directory */
+    int rv = fstat(fileno(m_dvi_fp), &fstatbuf);
+    if (rv == -1) {
+      perror("fstat");
+    }
+    if (S_ISDIR(fstatbuf.st_mode)) {	/* if it's a directory */
 	*errflag = FILE_IS_DIRECTORY;
 	fclose(m_dvi_fp);
 	m_dvi_fp = NULL;


which I think is correct (well, probably calling perror(3) isn't how xdvi wants to handle errors, but close enough for debuggin),
and it did not change anything. xdvi still crashes after shift-R.

That said, if insert a return True; after the fstat() call, I do see Dicks' crash prior to hitting shift-R.
So something depends on fstat() being called here, which is...weird! (As previously noted.)

--
jhawk at alum.mit.edu
John Hawkinson