Presently, if the argument to -r is a FIFO, ethereal and tetheral
crash. It musn't be a FIFO because (t)ethereal wants to seek on it; the
attached patch to wiretap/file.c prevents the user specifying a FIFO.
It turns out there's a more fundamental reason for the crash, though.
There's a bug in zlib 1.1.3 and 1.1.4 (and presumably earlier versions
too) where gzseek() doesn't save its internal error status if its call
to fseek() fails. This means a subsequent call to gzerror() sets errnum
to whatever the error was before the call to gzseek(), which is likely
to be Z_OK (gzopen() and gzdopen() initialise the error status to
Z_OK). The open_cap_file() routines in file.c and tethereal.c get
fooled and pass back an error status of 0 to their callers if
wtap_open_offline() fails in its call to file_seek(). Attached are
patches to these two files. The patch to file_wrappers.c merely
documents the problem.
I reported the problem to the zlib developers yesterday, but I haven't
heard back from them yet.
No doubt there could be more complicated workarounds, but as far as I
can see with a quick look, there are no other places where Ethereal gets
caught badly by the zlib problem -- the worst that could happen is that
the user receives an error message saying "error: success". How likely
is an error in fseek() with a regular file anyway? The open_cap_file()
patches have the advantage that the functions don't need to be changed
when zlib gets fixed.
Graeme Hewson
--- file.c.orig Thu May 23 11:27:12 2002
+++ file.c Wed Jun 5 19:07:17 2002
@@ -198,7 +198,13 @@
fail:
simple_dialog(ESD_TYPE_CRIT, NULL,
file_open_error_message(err, FALSE), fname);
- return (err);
+/*
+ * Due to a bug in zlib (1.1.4 is the current release as of 5 June 2002;
+ * 1.1.5 is not yet available), err can be 0 if the call to file_seek in
+ * wtap_open_offline() failed. Routines which call us only care if the
+ * return value is zero or non-zero, so return 1 on failure instead of err.
+ */
+ return (1);
}
/* Reset everything to a pristine state */
--- file_wrappers.c.orig Wed Feb 6 09:58:30 2002
+++ file_wrappers.c Wed Jun 5 19:55:51 2002
@@ -131,6 +131,18 @@
switch (errnum) {
case Z_OK: /* no error */
+/*
+ * There's a bug in zlib 1.1.4 (and presumably earlier versions too)
+ * where gzseek() doesn't save its internal error status if its call
+ * to fseek() fails. This means the subsequent call to gzerror() sets
+ * errnum to whatever the error was before the call to gzseek(), which
+ * is likely to be Z_OK. We don't work around the bug here, but leave
+ * it to the caller not to be fooled by file_error() returning 0 after
+ * a failed call to file_seek().
+ *
+ * There's a similar bug in gzrewind(), but as of Ethereal 0.9.4 there
+ * are no calls to this function.
+ */
return 0;
case Z_STREAM_END: /* EOF - not an error */
--- tethereal.c.orig Thu May 23 00:22:55 2002
+++ tethereal.c Wed Jun 5 19:08:22 2002
@@ -1818,5 +1818,11 @@
fail:
snprintf(err_msg, sizeof err_msg, file_open_error_message(err, FALSE), fname);
fprintf(stderr, "tethereal: %s\n", err_msg);
- return (err);
+/*
+ * Due to a bug in zlib (1.1.4 is the current release as of 5 June 2002;
+ * 1.1.5 is not yet available), err can be 0 if the call to file_seek in
+ * wtap_open_offline() failed. Our caller only cares if the return value
+ * is zero or non-zero, so return 1 on failure instead of err.
+ */
+ return (1);
}
--- wiretap/file.c.orig Wed May 29 03:19:49 2002
+++ wiretap/file.c Wed Jun 5 19:14:14 2002
@@ -181,7 +181,7 @@
*err = errno;
return NULL;
}
- if (! S_ISREG(statb.st_mode) && ! S_ISFIFO(statb.st_mode)) {
+ if (! S_ISREG(statb.st_mode)) {
if (S_ISDIR(statb.st_mode))
*err = EISDIR;
else
