summaryrefslogtreecommitdiff
path: root/paste/httpserver.py
Commit message (Collapse)AuthorAgeFilesLines
* Fix ticket #249maluke2008-04-131-7/+7
|
* allow binding to port 0pjenvey2008-03-071-1/+1
|
* Typoianb2008-02-191-1/+1
|
* Do not ignore wsgi.input readline size argument (which can lead to lots of ↵ianb2008-02-121-1/+4
| | | | memory used)
* Use threadsafe connection class, to avoid some https errors (#211)ianb2007-12-171-2/+9
|
* don't lower-case headers going out, send them as-iscce2007-11-081-3/+3
|
* minor tweak (appears to be backward compatible) to permit use with SSL chainscce2007-09-191-1/+1
|
* reverting FreeBSD threadpool disablingianb2007-09-041-5/+1
|
* fix more possible thread errors, followup to r6843pjenvey2007-08-131-5/+8
|
* Fix possible thread error, when a thread finishes while add_task is looking ↵ianb2007-08-131-2/+2
| | | | at it
* being pedantic, trying to find source of memory leakcce2007-08-091-0/+1
|
* The threadpool code just doesn't work on FreeBSD.cce2007-08-041-1/+7
|
* Don't read past the end of the line with LimitedLengthFile.readline (from ↵ianb2007-06-281-3/+1
| | | | Michael van Tellingen)
* docstring misformatting; added news linkianb2007-06-261-1/+1
|
* Cleaned up the worker lifecycle logic, fixing some bugs in how workers are ↵ianb2007-04-271-34/+48
| | | | tracked. Track why worker threads are started, and include this information in the logs.
* Fix the way max_requests works, so that we don't think the idle worker ↵ianb2007-04-261-1/+7
| | | | thread is still around (which caused the *real* worker threads to be culled, leaving only ghost workers who don't actually do any work, and thus the server would freeze)
* Add a better logging message about the thread worker pollingianb2007-04-261-0/+23
|
* Don't register the cleanup until the rest of the threadpool constructor has ↵ianb2007-04-261-3/+2
| | | | successfully run
* Add information about proxy requestsianb2007-04-151-1/+5
|
* Added a parameter to the threadpool: max_requests, which is the maximum ↵ianb2007-04-121-12/+25
| | | | number of requests to process in a worker thread before killing that thread. This should resolve problems with long-lived threads, as no thread lives too terribly long with this on. Also, turn the threadpool back on by default. And turn LimitedLengthFile back on when it is not an SSL connection
* - commeting out LimitedLengthFilecce2007-03-291-2/+10
| | | | | | | - making threadpool not be the default Once issues with these can be fixed, we could restore LimitedLengthFile and perhaps make threadpool the default
* Fixed logic error where we'd spawn extra workers, then immediately cull ↵ianb2007-03-291-1/+7
| | | | them. Now we wait a little while after spawning workers to cull them
* adding comment about address_string()cce2007-03-271-0/+3
|
* I think one meant min, not max here; if the read function askedcce2007-03-211-1/+2
| | | | | | for 16 bytes, and 64 are available, you should only return 16, not 64. This is important due to MIME types (where sub-chunks are read in via content-length on each sub-item).
* Give a better error when start_response isn't called; give a Content-Length ↵ianb2007-03-201-2/+8
| | | | so that the connection is closed properly on error
* fix some typos, from pythyianb2007-03-171-3/+3
|
* Changed default for kill_thread_limit; reformatianb2007-03-121-10/+11
|
* Don't start workers until everything is initializedianb2007-03-111-2/+2
|
* Major threadpool features to add threads temporarily, kill threads, monitor ↵ianb2007-03-101-25/+494
| | | | the killed threads, and notify the administrator about problems. Also updates to watchthreads to see this new information.
* removing outdated commentcce2007-03-071-2/+1
|
* - keep lookups as default, since that's how it was beforecce2007-03-051-1/+1
|
* making host address lookups optionalcce2007-03-051-5/+13
| | | | not looking up host addresses if they are private NATs
* Call exc_clear after worker threads are done, to remove any lingering frames ↵ianb2007-02-211-1/+12
| | | | from past exceptions
* Do not set REMOTE_HOST, as it requires a reverse DNS lookupianb2007-02-051-1/+11
|
* Allow httpserver's readline to take an optional argument (max length) which ↵ianb2007-02-011-2/+2
| | | | the cgi module uses.
* Limit number of bytes returned by environ['wsgi.input'].read(), so you can't ↵ianb2007-01-301-0/+63
| | | | read past end
* make paste.httpserver support all request methodsianb2007-01-301-2/+16
|
* Allow killing of errant threads through the web application (also add a ↵ianb2007-01-301-8/+39
| | | | module to kill threads with ctypes, and a method to kill threads from the httpserver thread pool)
* Add a variable to track the requests currently being processed by the ↵ianb2007-01-301-3/+18
| | | | httpserver (to see wedged threads)
* normpath returns '/' for the root, avoid reapplying the end slash in that casepjenvey2007-01-061-1/+1
|
* Not that good fixup of a problem where the end slash is lost under Paste's ↵ianb2007-01-051-0/+4
| | | | HTTP server
* SocketServer doesn't like new-style request handler classespjenvey2007-01-051-1/+1
|
* convert old-style classes to new-style classespjenvey2007-01-051-2/+2
|
* time.time() is too coarse for cert serials: use a randint insteadpjenvey2006-12-201-1/+1
| | | | (thanks jek)
* daemon_threads fix in r6009 was invalid, revertingpjenvey2006-12-201-3/+2
|
* o the _auto_ssl_context cert serial number must be an int not a longpjenvey2006-12-201-6/+10
| | | | | o show the correct protocol in serve's advertised URL o allow serve's daemon_threads option with WSGIThreadPoolServer
* Don't confuse people with inaccurate messageianb2006-12-201-2/+3
|
* Add atexit importianb2006-12-181-1/+2
|
* Apply #146, avoid orphaned threadpool, from jekianb2006-12-181-6/+12
|
* Apply #145, put http:// before URLs when starting serverianb2006-12-181-1/+4
|
View Lorry Controller Status