aw@bae.bae.bellcore.com (Andrew Wason) (12/12/90)
Since I installed fix-18, xmh often core dumps in the following situation: Select a message in the Toc Select "Edit Message" from the "View" menu Perform a simple editing command in the message Select "Save Message" from the "View" menu Select "Edit Message" from the "View" menu Perform a simple editing command in the message Select "Save Message" from the "View" menu Xmh usually core dumps at this point. The stack trace is: CoreSetValues() at 0xf76f53bc CallSetValues() at 0xf7708174 CallSetValues() at 0xf7708144 CallSetValues() at 0xf7708144 CallSetValues() at 0xf7708144 XtSetValues() at 0xf7708484 DoSaveView() at 0x10c00 _XtCallCallbacks() at 0xf76f0998 XtCallCallbacks() at 0xf76f0ba0 `Sme`Notify() at 0xf777d3b0 `SimpleMenu`Notify() at 0xf777c898 _XtTranslateEvent() at 0xf770fe10 DispatchEvent() at 0xf76f8ee8 DecideToDispatch() at 0xf76f9618 XtDispatchEvent() at 0xf76f96d8 main() at 0x5594 This looks like the same problem xrn 6.13 has. Both xmh and xrn do a get_values on a widgets XtNtranslations resource. Xmh does a get_values on the view widgets XtNtranslations and saves this translation table, then it installs some accelerators and does another get_values to get the modified translation table which it also saves. Then when the user switches between edit and read mode, xmh does a set_values on one or the other of the previously saved translation tables. Xrn does something similar to this. The problem is, as of fix-18 widgets appear to modify their old translation table when a new one is installed. This invalidates xmh's saved translation tables, and so it eventually does a set_values with one of these bad tables. Is this a bug in the Intrinsics/Athena widgets or xmh? Should widgets return a copy of their translation table on a get_values (that would fix xmh/xrn)? Should applications treat XtNtranslations as a private widget resource and never do a get_values on it (since the widget could modify it at any time behind the applications back)? Thanks for any input, Andrew [ Sun Sparc1+, SunOS 4.1, MIT X11R4 fix-18 ] -------------------------------------------------------------------------------- Andrew Wason Bell Communications Research aw@cellar.bae.bellcore.com Piscataway, NJ bellcore!cellar!aw
vania@osf.ORG (Vania Joloboff) (05/03/91)
I am running what I think is the the latest and greatest xmh R4 version, compiled with gcc on a DEC 3100 using ULTRIX 4.1 Regularly, I observe the following, but I can't reproduce it systematically, so I can't really send a bug report to xbugs... When I reply to a mail and I insert the replied mail, using the xmh insert command, then if I use button3 to select some text from the current cursor position (without any previous button action) xmh goes into an infinite loop, and the only way out is to kill the process. Has anybody observed the same behavior ? -- Vania