Since I switched to Squid 1.1, HTML error messages apparently get a null
byte (\0) at the end. This confuses some programs (like UCB-Mail) when
doing "command-line" retrievals.
I do not observe this with Squid 1.0 machines, neither with normal HTML
files, only when there is a cache-detected failure retrieving the file
(connection refused, DNS error, etc). You can use a command like that one
to see the spurious character:
url_get -d http://wrong.server/ | od -c
Instrumenting the source code, I can see the error message which is
correct, even when stored in the entry with storeAppend (in store.c). It
is after storeAppend that it becomes corrupted.
Received on Thu Dec 19 1996 - 03:29:08 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:33:55 MST