R devel - Oct 2007 - (PR#10379) Re: x11(....) kills R without DISPLAY

Hin-Tak Leung wrote:
> Peter Dalgaard wrote:
> <snipped>
>> You need x11() with a valid display to trigger the bug:
>>
>> [pd at titmouse2 BUILD]$ ssh -Y 192.168.1.10
>> pd at 192.168.1.10's password:
>> Last login: Sat Oct 27 02:40:16 2007 from 192.168.1.11
>> [pd at janus ~]$ echo $DISPLAY
>> localhost:10.0
>> [pd at janus ~]$ DISPLAY=3D R -q
>>  > x11("localhost:10.0")
>> Error: Couldn't find per display information
>> [pd at janus ~]$ uname -a
>> Linux janus 2.6.22.9-91.fc7 #1 SMP Thu Sep 27 20:47:39 EDT 2007=20
>> x86_64 x86_64 x86_64 GNU/Linux
>> [pd at janus ~]$ cat /etc/issue
>> Fedora release 7 (Moonshine)
>> Kernel \r on an \m
>
> Agh, sorry. Yes, x11() (with or without $DISPLAY set) doesn't
> die catatrophically, x11("validinfo") does.
>
> HTL
The culprit would seem to be this bit of devX11.c

1302                    xtdpy =3D XtOpenDisplay(app_con, NULL,
"r_x11",=20
"R_x11",
1303                                          NULL, 0, &zero, NULL);
1304                    toplevel =3D XtAppCreateShell(NULL, "R_x11",

The 2nd arg to XtOpenDisplay is listed as display_string, so passing a=20
NULL here seems like trouble when the default ways of finding the=20
display do not work.

Looks like a fix is to insert p instead of NULL. (Tested rudimentarily.)



--=20
   O__  ---- Peter Dalgaard             =D8ster Farimagsgade 5, Entr.B
  c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
 (*) \(*) -- University of Copenhagen   Denmark          Ph:  (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)                  FAX: (+45) 35327907

On 10/29/07, Prof Brian Ripley <ripley@stats.ox.ac.uk>
wrote:
>
> The problem is that the XtOpenDisplay call did not specify the display.
> Easily fixed ....
>
On 10/29/07, p.dalgaard@biostat.ku.dk <p.dalgaard@biostat.ku.dk>
wrote:
>
>
> The culprit would seem to be this bit of devX11.c
>
> 1302                    xtdpy = XtOpenDisplay(app_con, NULL,
"r_x11",
> "R_x11",
> 1303                                          NULL, 0, &zero, NULL);
> 1304                    toplevel = XtAppCreateShell(NULL,
"R_x11",
>
> The 2nd arg to XtOpenDisplay is listed as display_string, so passing a
> NULL here seems like trouble when the default ways of finding the
> display do not work.
>
> Looks like a fix is to insert p instead of NULL. (Tested rudimentarily.)

Prof. Ripley replaced that NULL with dsp -- p is normally the same.  It's in
svn (r43300 on he trunk, r43301 on R-2-6-branch).
It built without a hitch, and the fix solves my problem.

P. Dalgaard also pinpointed the problem.  And confirmed Xt was involved :-)
Apparently an XOpenDisplay on the specified display may be followed by an
XtOpenDisplay, previously on the default display.

Thanks to everyone who helped figure this out and fix the problem!
/Christian Brechbuehler

	[[alternative HTML version deleted]]