wessels [Sat, 15 Nov 1997 00:22:09 +0000 (00:22 +0000)]
Fix storeClientCopy2() bug. We can't compare copy_offset to
inmem_hi because for inmem_hi == 0 for swapped out objects. Must
compare object_len instead!
wessels [Sat, 15 Nov 1997 00:20:13 +0000 (00:20 +0000)]
In clientReadRequest(), moved reinstalling the read handler with
commSetSelect() to just after the read(). Due to another bug, there
was a problem in clientReadRequest that we called commSetSelect()
on a FD which had been closed.
wessels [Fri, 14 Nov 1997 23:05:48 +0000 (23:05 +0000)]
Fix 'EOF' bug in storeClientCopy2(). If the client copy offset equals
inmem_hi, and store_status == STORE_OK, then callback with zero size
to indicate EOF.
wessels [Thu, 13 Nov 1997 06:36:19 +0000 (06:36 +0000)]
Ron Gomes fixes.
- Remove USE_PROXY_AUTH and LOG_FULL_HEADERS from Makefile.in
- Fixed log_full_hdrs bug
We can't use pathname_stat() for both default_all() and parsing the
config file. We must check the pathnames only after both the
defaults and the config file have been parsed. DW also made misc
other fixes to cache_cf.c and friends.
wessels [Tue, 11 Nov 1997 04:32:00 +0000 (04:32 +0000)]
OOF! Somehow this bit of logic got lost which releases objects
if the request had an Authorization header, but the reply did NOT
have a Proxy-Revalidate header.
wessels [Tue, 11 Nov 1997 03:54:30 +0000 (03:54 +0000)]
Bad things were happening when we copied the mem->reply structure
from a 'NOT MODIFIED' reply over to the old object. Specifically,
www.thegist.com (netscape/1.13) sends us a bogus content-length.
this content length seems to refer to the size of the NOT MODIFIED
reply headers instead of the URI body!
wessels [Mon, 3 Nov 1997 23:19:23 +0000 (23:19 +0000)]
From: Henrik Nordstrom <hno@hem.passagen.se>
* Allow %escapes to be send in a FTP url. The path is now handled correctly:
1) Split the path on /
2) Decode each part (possibly introducing / or ..)
* Removed up the "cleanup" of URL's.. it is basically not a good idea to
try to clean up the URL. If we get a strange URL it is probably the
intention. The "cleanup" code usually makes a mess of things.. And when
this cleanup was removed, it was much easier to handle "BASE HREF" or
not...
* Added back .. in listings (code was not updated)
* Changed .. to generate %2e%2e/ when appropriate ("root" directory, or
after %2e%2e)
* Added a link to %2f/ when on "Top level" URL.
* Added a "Up" link to ../ on the above two ones. (I chosed the term Up,
since it is to a shorter URL.. whether this it up/down in the file
system is hard to tell...)
* Changed text to "Home directory" when in directory /%2f/
wessels [Mon, 3 Nov 1997 23:05:23 +0000 (23:05 +0000)]
From: Henrik Nordstrom <hno@hem.passagen.se>
Here is a small patch to squid-1.2.beta4 to support %nn escapes in FTP
URLs... (all code was already there but never called...)