diff options
Diffstat (limited to 'src/cleanup/bug1.in')
| -rw-r--r-- | src/cleanup/bug1.in | 41 |
1 files changed, 41 insertions, 0 deletions
diff --git a/src/cleanup/bug1.in b/src/cleanup/bug1.in new file mode 100644 index 0000000..bda18cf --- /dev/null +++ b/src/cleanup/bug1.in @@ -0,0 +1,41 @@ +#verbose on +open bug1.file.tmp + +# Symptom: +# +# infinite loop in postcat and in delivery agents +# +# Cause: +# +# Failure to update location info after following a pointer record, +# while updating a message header record +# +# Analysis: +# +# This happens with repeated updates of the same message header. +# After the first update, the update #1 header record sits in the +# heap at the end of the queue file, and is followed by a reverse +# pointer to the start of the next message header record or the +# message body, somewhere in the middle of the queue file. +# +# The problem started with update #2 of that same message header. +# While following the reverse pointer record after the update #1 +# header record to find out the start of the next header or message +# body, the header updating routine did not update its notion of +# where it was. Thus, it believed that the next header or body record +# was located after the reverse pointer record. That was not the +# middle of the message, but the end of the queue file. The second +# update would result in an update #2 header record, followed by a +# reverse pointer to what used to be the end of the queue file, but +# had meanwhile become the location of the update #2 header record. +# +# Thus, anything that tried to deliver mail would loop on the update +# #2 header record. After update update #3 of the same header, the +# delivery agent would loop on the update #3 record, etc. + +upd_header 1 Subject long header text +upd_header 1 Subject long header text +upd_header 1 Subject long header text +upd_header 1 Subject long header text + +close |