Bug#414307: xnest: Crashes with client Tulip

[Brice Goglin <Brice.Goglin@ens-lyon.org>]

Version: 2:1.2.99.901-1


J.P. Larocque wrote:
> Xnest consistently crashes with the application Tulip (Debian package
> tulip, version 2.0.6-4).  Simply create a new document.
>   

I can reproduce the problem here, even with Xnest currently in
experimental. Here's a backtrace from a debug-enabled Xnest 2:1.2.99.901-1:

Program received signal SIGSEGV, Segmentation fault.
0x081d8b7d in _mesa_update_draw_buffer_bounds (ctx=0x8f9b1b8) at
framebuffer.c:393
393        if (buffer->Name) {
(gdb) bt full
#0  0x081d8b7d in _mesa_update_draw_buffer_bounds (ctx=0x8f9b1b8) at
framebuffer.c:393
        buffer = (struct gl_framebuffer *) 0xb7684008
#1  0x081d8a7e in _mesa_resize_framebuffer (ctx=0x8f9b1b8,
fb=0xb7652008, width=100, height=30)
    at framebuffer.c:332
        i = 19
#2  0x08148a43 in xmesa_check_and_update_buffer_size (xmctx=0x8f9b1b8,
drawBuffer=0xb7652008)
    at xm_api.c:1890
        width = 100
        height = 30
#3  0x08148b35 in XMesaMakeCurrent2 (c=0x8f9b1b8, drawBuffer=0xb7652008,
readBuffer=0xb7652008)
    at xm_api.c:1940
No locals.
#4  0x081451f8 in __glXMesaContextMakeCurrent (baseContext=0x8f993f8) at
../../../../GL/mesa/X/xf86glx.c:200
        context = (__GLXMESAcontext *) 0x8f993f8
        drawPriv = (__GLXMESAdrawable *) 0x96675c0
        readPriv = (__GLXMESAdrawable *) 0x96675c0
#5  0x08120076 in DoMakeCurrent (cl=0x86e5068, drawId=2099970,
readId=2099970, contextId=2099967, tag=1)
    at ../../../GL/glx/glxcmds.c:662
        client = (ClientPtr) 0x86e7558
        reply = {type = 1 '\001', unused = 0 '\0', sequenceNumber = 0,
length = 2, contextTag = 150354040,
  pad2 = 150353544, pad3 = 3213761352, pad4 = 135071716, pad5 =
139016604, pad6 = 3213761156}
        drawPixmap = (__GLXpixmap *) 0x0
        readPixmap = (__GLXpixmap *) 0x0
        glxc = (__GLXcontext *) 0x8f993f8
        prevglxc = (__GLXcontext *) 0x8f993f8
        drawPriv = (__GLXdrawable *) 0x96675c0
        readPriv = (__GLXdrawable *) 0x96675c0
        error = -1081206120
        mask = 0
#6  0x0811facf in __glXDisp_MakeCurrent (cl=0x86e5068, pc=0xb7945290
"\221\005\004")
    at ../../../GL/glx/glxcmds.c:408
        req = (xGLXMakeCurrentReq *) 0xb7945290
#7  0x080ffcab in __glXDispatch (client=0x86e7558) at
../../../GL/glx/glxext.c:551
        stuff = (xGLXSingleReq *) 0xb7945290
        opcode = 5 '\005'
        proc = (__GLXdispatchSingleProcPtr) 0x811fa83
<__glXDisp_MakeCurrent>
        cl = (__GLXclientState *) 0x86e5068
        retval = -1081206112
#8  0x080d09d4 in XaceCatchExtProc (client=0x86e7558) at
../../Xext/xace.c:299
        stuff = (xReq *) 0xb7684008
        major = 145
        ext = (ExtensionEntry *) 0x84bba68
#9  0x0806ff93 in Dispatch () at ../../dix/dispatch.c:457
        clientReady = (int *) 0xbf8e1b70
        result = 16
        client = (ClientPtr) 0x86e7558
        nready = 0
        icheck = (HWEventQueuePtr *) 0x84a21c4
        start_tick = 4240
#10 0x0808ec5e in main (argc=2, argv=0xbf8e2094, envp=0xbf8e20a0) at
../../dix/main.c:477
        i = 1
        j = 2
        k = 2
        error = 138905476
        xauthfile = 0xbf8e3fba "/home/bgoglin/.Xauthority"
        alwaysCheckForInput = {0, 1}

As a workaround, you might want to try Xephyr (from xserver-xephyr)
which is the recommended replacement for Xnest nowadays. I doesn't crash
under the same circumstances here.

I wonder whether we should stop building Xnest and add a transitional
package to Xephyr. There are multiple Xnest bugs that don't show up in
Xephyr...

Brice