Commit by micah :: r5956 inputpipe/Makefile:

Whoops, use $(CC)

Commit by micah :: r5954 inputpipe/uinput/README:

Let people know that they don't need this in linux-2.6.10-rc2-mm4 or later

Commit by micah :: r5880 inputpipe/README:

Mention building and installing just a bit, maybe

Commit by micah :: r5879 inputpipe/ (COPYING README):

This Might Be a Readme

Commit by micah :: r5878 inputpipe/Makefile:

Some craaaaazy people might want "make install"

Commit by micah :: r5871 inputpipe/src/input.h:

Superfluous headers! die!

Commit by micah :: r5870 inputpipe/uinput/ (input.h uinput.c uinput.h uinput.txt):

Move EV_UINPUT to input.h and the long comment from uinput.h into a separate text file. This is per Aristeu's suggestions.

Commit by micah :: r5869 inputpipe/uinput/input.h:

Add an unmodified input.h from Linux 2.6.10-rc1

Commit by micah :: r5868 inputpipe/uinput/uinput.c:

Apply Andrew Morton's patch to remove unnecessary casts when assigning to and from void*'s.

Commit by micah :: r5715 inputpipe/src/server.c:

Set a default physical address, so even if the client never sends us one we still get to label our input device with the client's IP address and port.

Commit by micah :: r5714 inputpipe/src/server.c:

Server-side artificial EV_SYN generation when necessary

Commit by micah :: r5713 inputpipe/src/client.c:

Detect the EV_SYN bit, telling us whether the kernel plans on generating synchronization events. If so, flush the write buffer after every event. If not, flush only after a sync. Removed client-side generation of artificial sync events- that can be done server-side now, so it's a waste of bandwidth.

Commit by micah :: r5712 inputpipe/ (5 files in 2 dirs):

Use a local copy of input.h, so we don't need all these workarounds for machines with old copies of it

Commit by micah :: r5711 inputpipe/src/ (client.c packet.c packet.h server.c):

A metric handful of fixes that unbreak our inetd support. The packet_socket object now supports separate fds for reading and writing, so we use stdin and stdout correctly. We also suppress stderr messages when in inetd mode (same as the -q flag) since they will be sent over the socket as well, thereby screwing up our protocol nastily.

Commit by micah :: r5710 inputpipe/src/server.c:

Fix client addresses when inputpipe-server is running through inetd

Commit by micah :: r5709 inputpipe/uinput/ (uinput.c uinput.h):

Go back to using a switch statement rather than a > comparison to determine which ioctls are allowed while the device is created. We can't make assumptions about what _IOW and _IOR put in the high bits of the ioctl number.

Commit by micah :: r5708 inputpipe/ (5 files in 2 dirs):

Finish physical path support. uinput's UI_SET_PHYS seems to be working now, and inputpipe has a new packet type for sending the physical path. inputpipe-server sets up a path of the form ipipe://client:port/client-path, so it's easy to identify exactly where a device comes from, even when inputpipe is involved.

Commit by micah :: r5707 inputpipe/uinput/ (uinput.c uinput.h):

Add a new ioctl for setting the device's physical path.

Commit by micah :: r5704 inputpipe/src/client.c:

Plug a leak of 1 file descriptor per device every couple seconds when inputpipe-client is trying to connect to the server.

Commit by micah :: r5703 inputpipe/uinput/uinput.c:

oops