[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference bulova::decw_jan-89_to_nov-90

Title:DECWINDOWS 26-JAN-89 to 29-NOV-90
Notice:See 1639.0 for VMS V5.3 kit; 2043.0 for 5.4 IFT kit
Moderator:STAR::VATNE
Created:Mon Oct 30 1989
Last Modified:Mon Dec 31 1990
Last Successful Update:Fri Jun 06 1997
Number of topics:3726
Total number of notes:19516

2516.0. "XSelectAsyncInput" by OPHION::KLEE () Mon Mar 26 1990 20:23

    XSelectAsyncInput() is a DECwindows addition to Xlib.  It causes Xlib
    to call a user-specified callback function whenever a particular X
    event arrives.  The DECwindows Xlib manual apparently places no
    restrictions on what can occur in this callback function.  Are people
    trying to put Xlib functions in the callback?  I don't think either
    the VMS or the Ultrix Xlib will work properly in that case.  Would
    anyone be especially annoyed if I changed the manual to require that
    the callback not contain Xlib functions?  The MIT Xlib manual already
    contains such a restriction on IfEvent and error handling callbacks.
    
    Ken
    
T.RTitleUserPersonal
Name
DateLines
2516.1Please Don't Do It!STAR::KLEINSORGEFred Kleinsorge, VMS DevelopmentTue Mar 27 1990 12:2625
    WRONG!!!!!
    
    VAX/UISX absolutely, positively MUST be able to do Xlib calls from
    the AST.  This product is just entering FT2 and yes, indeed it does
    work.  There have been and still are a few bugs in Xlib and the
    transports, and a couple make the V1 server unreliable, but the
    server as of V5.3-1 is reliable (at least reliable enough for UISX).
    
    UISX layers the VWS (the old VMS windowing software) programming
    interface -- UIS -- on top of VMS/DECwindows.  UIS relies heavily on
    the doorbell AST mechanism and most applications are event driven,
    emitting Xlib calls generated by user code called as a result of the
    event.
    
    If Xlib cannot be called from the doorbell AST, then it's next to
    useless.  While there might be workarounds, they would be cumbersome
    and slow - and would make the product non-viable.  If the ability to
    do this is removed, or support for it decommitted, then you jepardize
    a large revenue stream for hardware depending on having the UIS
    application base available.
    
    I also believe that there are real customers using this in the same
    way.  Probably the customers that forced us to invent it to begin
    with.
    
2516.2FLUME::dikeTue Mar 27 1990 13:433
It will work right on Ultrix if you use Xlib within the callback.  I was very
careful about making sure that worked right.
				Jeff
2516.3DECWIN::FISHERPrune Juice: A Warrior's Drink!Tue Mar 27 1990 13:467
Indeed...that is the whole point.  Xlib is supposed to be reentrant from async
callbacks.  As someone said, there are a few bugs which will be fixed soon.

The TOOLKIT, however, is a different story.  You can't call any Xt calls from
an AST/Signal.

Burns
2516.4A view from the fieldSCAM::DIALTue Mar 27 1990 14:1311
    I just completed two months of work porting an application to Xlib for
    an OEM.  This product depends on the ability to do Xlib calls from an
    AST.  From my experience with this customer's application, porting a
    straight line application with AST driven keyboard input almost demands
    the XSelectAsyncInput and Xlib calls from AST's features.  Granted, it
    would have been preferable to restructure the application to use the
    toolkit, but this was not possible.  As more customer applications are
    converted to use DECwindows/X-windows, the impact of removing the
    feature could be dramatic.
    
    Barry
2516.5Major customer's ARE using thisSITBUL::MCCARTNEYTue Mar 27 1990 17:018
    RE: .1
    
    One of the major customers who required this (a rather large customer
    in Texas, also a CMP), does make wide use of this feature.  I'd hate to
    see what happens if we tell them now that we're pulling support.  They
    may carry through with their initial threat of calling Sun.
    
    Irene
2516.6Major portions of DECtp depend upon this working....IO::MCCARTNEYJames T. McCartney III - DTN 381-2244 ZK02-2/N24Tue Mar 27 1990 18:119

In fact, we are pushing for the tookit and Xlib to become THREAD-reentrant.
We can probably survive with just AST-reentrancy but it is an absolute 
requirement that we be able to call Xlib from the AST. When CMA comes along
fully-reentrant code will become even more important.

James McCartney
DECforms Development
2516.7KONING::KONINGNI1D @FN42eqWed Mar 28 1990 19:265
I agree... every major software toolset needs to be thread-reentrant,
all the various pieces of DECwindows very much among them.  ASTs will
have to do until threads show up.

	paul