~peterh/xgob - Trying a possible Go XCB design

Age	Commit message (Collapse)	Author	Files	Lines
2013-03-13	Update for Go 1HEAD master	Peter Harris	10	-69/+41
	go fix, shuffle directories for the go tool so we don't need a Makefile, and remove the Makefile.
2011-08-11	gofix for current release	Peter Harris	3	-13/+13

2011-03-02	Fix reading of errors from the server	Peter Harris	1	-1/+1
	All replies/events/errors would be broken after the first error, due to a short read.
2011-02-15	Add Auth support	Peter Harris	2	-7/+180
	MIT-MAGIC-COOKIE only so far. The XDM auth is considerably more complex.
2011-02-15	Add Get* Xau implementation	Peter Harris	1	-0/+200

2011-02-06	Cleanup: Automatic reformat	Peter Harris	7	-202/+206
	find . -name "*.go" \| xargs gofmt -w
2011-02-03	Update for latest go release: nonblocking channel read	Peter Harris	1	-2/+5

2011-01-21	Add license file	Peter Harris	1	-0/+31

2011-01-12	Quote command (like the C xlsclients)	Peter Harris	1	-3/+75

2011-01-11	Change xmu.go/ClientWindow to more closely match the C version	Peter Harris	1	-4/+3
	It's a completely cosmetic change, but it makes comparing the two versions slightly easier.
2011-01-11	Fix typo in TCP selection	Peter Harris	1	-1/+1
	Otherwise, we try to connect to remote hosts using Unix (local) sockets.
2011-01-07	xgob/xproto depends on xgob	Peter Harris	1	-0/+1

2011-01-07	Add (and use) callback-based reply decoders	Peter Harris	2	-12/+35
	This improves performance, as we aren't spinning up and killing a tonne of goroutines.
2011-01-07	Trim one round-trip out of xmu.ClientWindow	Peter Harris	1	-14/+16

2011-01-07	Compile sub-packages into the correct location	Peter Harris	4	-8/+9

2011-01-02	xlsclients now works	Peter Harris	6	-22/+431

2011-01-02	Thinking out-loud about what xproto might look like	Peter Harris	5	-1/+299

2010-12-28	Stop using defer when it isn't needed	Peter Harris	1	-5/+5
	It's a powerful construct, but it cuts the speed of noop.go in half. That's a steep penalty when none of the functions return (and none should panic). Now pushing 2.2M noop/sec
2010-12-28	Allow nil error channel in WriteNoreplyRequest	Peter Harris	2	-20/+21
	Not allocating a channel at all doubles the speed of the noop test. Now we're pushing just over 1M noop/sec (compared to x11perf/Xlib, which pushes 13M noop/sec)
2010-12-28	Pass channel into WriteNoreplyRequest	Peter Harris	2	-16/+34
	This speeds up noop.go 3x (or 1.5x with the GC disabled)
2010-12-28	Take write lock in Flush()	Peter Harris	1	-0/+3

2010-12-28	Buffer output	Peter Harris	1	-3/+9
	This roughly doubles the speed of the noop test app.
2010-12-24	Add simple noop test app	Peter Harris	3	-3/+45
	On my mac, x11perf does 13 million noop/sec This test app does 0.11 million noop/sec
2010-12-24	Remove extra channel send	Peter Harris	1	-2/+5
	This extra send causes deadlocks when the cookie is discarded.
2010-12-24	Automatically insert Sync requests where needed	Peter Harris	2	-4/+25

2010-12-24	Fix connecting to Unix sockets	Peter Harris	1	-1/+1

2010-12-24	Try harder to avoid deadlocks	Peter Harris	1	-6/+14
	By dropping the reply queue lock before actually trying to write, we reduce the chances that the server won't read our blocking socket due to us not reading events/replies/etc from earlier.
2010-12-24	Add simple Makefile	Peter Harris	2	-1/+17

2010-12-22	Basic functionality complete	Peter Harris	2	-18/+236
	test.go functions as expected if you specify the full display as tcp/localhost:0 and the server has xauth disabled.
2010-12-22	Read entire connection block	Peter Harris	1	-13/+44

2010-12-21	Initial commit	Peter Harris	3	-0/+213
	Work in progress. Nothing actually works yet.