xpra icon
Bug tracker and wiki

Opened 4 years ago

Closed 3 years ago

Last modified 3 years ago

#1136 closed enhancement (fixed)

move websockify in the xpra process

Reported by: Antoine Martin Owned by: alas
Priority: major Milestone: 1.0
Component: server Version: trunk
Keywords: Cc:

Description

So we don't have to exec anything, then we also won't be needing the proxy instance running xpra side (which is forwarding to and from the websockify process, using 2 threads).
See also #1134.

Attachments (2)

websockify-inprocess.patch (22.0 KB) - added by Antoine Martin 3 years ago.
move websockify handling in process
web-ssl.patch (10.9 KB) - added by Antoine Martin 3 years ago.
attempts at preventing ssl from deadlocking - failed

Download all attachments as: .zip

Change History (9)

comment:1 Changed 4 years ago by Antoine Martin

Milestone: 0.170.18
Status: newassigned

Changed 3 years ago by Antoine Martin

Attachment: websockify-inprocess.patch added

move websockify handling in process

comment:2 Changed 3 years ago by Antoine Martin

The patch above works fine, only left todo:

  • handle html=/path/to/www and generate / detect correct path (ie: on osx, in the app bundle)
  • TLS setup
  • man page updates
  • test tcp-proxy mode better
  • use the peek function to avoid creating a whole protocol instance (and threads) if the packet header is invalid
  • verify blocking sockets and timeout values
Last edited 3 years ago by Antoine Martin (previous) (diff)

comment:3 Changed 3 years ago by Antoine Martin

Mostly done in r12688. Firefox is still having problems, but that doesn't seem to be anything new.
Still todo:

  • TLS
  • bail out early on invalid packet header

Changed 3 years ago by Antoine Martin

Attachment: web-ssl.patch added

attempts at preventing ssl from deadlocking - failed

comment:4 Changed 3 years ago by Antoine Martin

Owner: changed from Antoine Martin to alas
Status: assignednew
  • r12708 fixes an error with win32 servers (but win32 servers still don't support the html5 client: #1211)
  • r12699 bails out early
  • I give up on SSL for now - will follow up in #1213, the patch above tries to use a thread as early as possible, setting timeouts on the sockets, raising the number of sockets we create, etc.. to no avail

@afarr: the html client should be faster now

comment:5 Changed 3 years ago by J. Max Mena

HTML client is by far faster. It's like night and day faster now.

I'd say performance in Chrome is 90% as good as Python.

comment:6 Changed 3 years ago by J. Max Mena

Resolution: fixed
Status: newclosed

Oh, sorry, I hit submit without thinking.

Closing.

comment:7 Changed 3 years ago by Antoine Martin

Milestone: 0.181.0

Milestone renamed

Note: See TracTickets for help on using tickets.