Changed rfc1035AnswersUnpack() to unpack DNS names into character
buffers so that the upper layers don't have to. This is the switch
statement inside that function that exectues different copies
based on the record type.
rfc1035AnswersUnpack would allocate an "array" of answers, even
if there were no answers in the reply, but there were other things
in the reply like auth and additional.
Henrik
This patch makes overloaded async-io a non-fatal error.
The patch does two things:
1. Stop caching objects if the async-io queue is way to large.
2. Sync async-io operations if async-io queue gets ridicolously large
(was a fatal error).
henrik sez: The time zone in common log format is miscalculated at
the end of the year (any year, not a y2k specific problem even if
next time it is visible is y2k..).
If you install Squid listening in a secondary IP address instead
of the primary of the interface, it will generate for FTP listings
pages that refer icons to the primary. Given that Squid is not
listening on that interface, it results in a long timeout before
the page displays and "broken graphics" My suggestion is that
visible_hostname default to whatever tcp_incoming_address is,
instead of defaulting to the canonical name of the machine.
From: Henrik Nordstrom <hno@hem.passagen.se>
<a patch that does a reverse lookup (BLOCKING!!) if tcp_incoming_address
is set. The answer is used as cache hostname if the lookup succeeds.>
rousskov [Wed, 24 Mar 1999 11:16:08 +0000 (11:16 +0000)]
- added calculation of Content-Length for multi-range responses;
the implementation packs HTTP headers into a buffer to calculate
their length; not very efficient, but simple.
wessels [Sat, 30 Jan 1999 05:17:09 +0000 (05:17 +0000)]
reconfigure caused peer address lookups to take a long time because
we didn't queue another peerRefreshDNS event. instead we waited for
the current, pending event, no matter how far off it was.
wessels [Sat, 30 Jan 1999 04:01:06 +0000 (04:01 +0000)]
we need to lock entry in errorAppend(). our storeAppend (or rather
storeBufferFlush) call might result in the client-side unregistering
and aborting, and unlocking, and freeing the entry. This happened
for netdb exchange requests.
wessels [Sat, 30 Jan 1999 00:36:37 +0000 (00:36 +0000)]
storeSwapOutFileOpened: Unable to open swapfile:
/disk2/cache2/00/71/0000716C (13) Permission denied
* yuck. don't clear the filemap bit for some errors so that
* we don't try re-using it over and over
wessels [Sat, 30 Jan 1999 00:25:08 +0000 (00:25 +0000)]
From: "Erick C. Chang" <erick@mail.thu.edu.tw>
I've corrected some mistakes and make some changes
on the Traditional Chinese(Big5) error messages that
come with squid 2.2 pre1(the old one I create).
Maybe you should use this new one when the next
squid release. This one is better than last one :p
wessels [Fri, 22 Jan 1999 06:58:41 +0000 (06:58 +0000)]
From: Henrik Nordstrom <hno@hem.passagen.se>
Here is a patch that replaces persistent_client_posts all together.
Instead it uses a (fast) ACL to determine if this extra CRLF should be
appended to the body of the request or not.
wessels [Fri, 22 Jan 1999 06:15:35 +0000 (06:15 +0000)]
From: Henrik Nordstrom <hno@hem.passagen.se>
This patch removes the last traces of cache_mem high/low water marks.
Keep hot object store at cache_mem instead of cache_mem_high.
projects / thirdparty / squid.git / log
