Re: [Nmh-workers] mhshow -file /tmp (!)

[David Levine]

Ken wrote:

> maybe the "right" fix is to properly bubble up read errors from m_getfld()?

I agree, and would be willing to visit that after 1.7.  I think that all callers
should be able to handle the error, but I'm not certain.

In the meantime, how about the fix below to parse_mime()?  That should be done
in any case, because parse_mime() really expects a file.

David

index 5db4efe..bd2b16c 100644
--- a/uip/mhparse.c
+++ b/uip/mhparse.c
@@ -211,0 +212 @@ parse_mime (char *file)
+    struct stat statbuf;
@@ -247,3 +248,12 @@ parse_mime (char *file)
-    } else if ((fp = fopen (file, "r")) == NULL) {
-       advise (file, "unable to read");
-       return NULL;
+    } else {
+        /* Don't try to parse a directory. */
+        if (lstat (file, &statbuf) == NOTOK) {
+            advise (file, "unable to lstat");
+            return NULL;
+        } else if (S_ISDIR(statbuf.st_mode)) {
+            advise (NULL, "%s is a directory", file);
+            return NULL;
+        } else if ((fp = fopen (file, "r")) == NULL) {
+            advise (file, "unable to read");
+            return NULL;
+        }