From: epg <>
Date: Thu, 16 Jan 2003 20:40:57 +0000 (+0000)
Subject: (strip_from): I added a comment about my confidence in the
X-Git-Url: https://diplodocus.org/git/mdeliver/commitdiff_plain/5ca5cba2fc37eb992e9fc173aaccc92576479c52?ds=inline;hp=cbab192ba54fdeb51f62f514cc6e5eeba9b4e98f

(strip_from): I added a comment about my confidence in the
read-first-five-bytes or fail bit a long time ago, but never
committed it.
---

diff --git a/mdeliver.c b/mdeliver.c
index 54884e9..f673597 100644
--- a/mdeliver.c
+++ b/mdeliver.c
@@ -153,6 +153,14 @@ strip_from(int fd, char *buf, char *fn)
 {
     ssize_t r;
 
+    /* Yes, read(2) can be interrupted and read fewer bytes than
+     * requested.  However, i'm only asking for five bytes here.  I
+     * have tested this program extensively and it handles large
+     * volumes of mail correctly every day on my systems.  This holds
+     * up my original opinion that expecting to be able to get five
+     * bytes safely is not an unreasonable demand.  If i can't get
+     * five measly bytes, something is probably wrong, so bomb.
+     */
     r = read(STDIN_FILENO, buf, 5);
     if (r != 5) {
         save_errno( unlink(fn) );