On Wed, 22 Sep 2004, Mark Vojkovich wrote:On Wed, 22 Sep 2004, Marc Aurele La France wrote:On Mon, 20 Sep 2004, Phil Dibowitz wrote:On Mon, Sep 20, 2004 at 03:13:40PM -0600, Marc Aurele La France wrote:Perhaps we can nail down what firefox/mozilla is repeatedly requesting.
What does `xdpyinfo -queryExtensions` say?
I'm assuming this wouldn't change when X freezes? Under that assumption, this is from X not frozen:
Thanks, this helps muchly. The only reason I can think of that this output would not correspond to that of the looping server would be extensions being initialised in a different order, a situation I see as highly unlikely, if at all possible.
On the assumption that this xdpyinfo output is in fact relevent, then your server strace shows, among other things, two client connections repeatedly calling XShmPutImage(), to which the server never replies (with success or failure).
At this point, I don't really know whether it's relevent to the problem that there are two such connections, nor whether the lack of server response is normal XShmPutImage() behaviour for this case. For the latter, I'd like you to capture another server strace, this one without Xinerama and as near as possible to the scenario that would cause the Xinerama loop. Please take care to not delete the original strace's you've made available through your web site.
If this non-hang strace shows a server response to the client's XShmPutImage() (or some other server behaviour), then there's likely something amiss with xc/programs/Xserver/Xext/shm.c's ProcPanoramiXShmPutImage(), perhaps with its handling of sendEvent, at which point I'd ask Mark, one with more knowledge of Xinerama internals than me, to investigate further.
My Xinerama system stays up for weeks on end and I spend alot of time using Mozilla and Netscape. Perhaps one of these new apps is using RENDER and it's something related to that? If there was a fundamental bug in shm's Xinerama support I would have expected somebody to find it long ago. That code is many years old and I'm not aware of any outstanding bugs. It seems more likely that there is some new extension usage going on, and that extension has some Xinerama issues.
The strace doesn't show any other extension being involved. IOW, there is no call to feel defensive about this. The facts will speak for themselves.
I think I might have located the problem. The server is not reporting the correct drawable in a completion event requested by an Xinerama-fied XShmPutImage request. The attached is a possible fix for that.
Marc.
+----------------------------------+-----------------------------------+ | Marc Aurele La France | work: 1-780-492-9310 | | Computing and Network Services | fax: 1-780-492-1729 | | 352 General Services Building | email: [EMAIL PROTECTED] | | University of Alberta +-----------------------------------+ | Edmonton, Alberta | | | T6G 2H1 | Standard disclaimers apply | | CANADA | | +----------------------------------+-----------------------------------+ XFree86 developer and VP. ATI driver and X server internals.
shm.diff.gz
Description: Binary data
