From:	MERC::"uunet!WKUVX1.BITNET!MacroMan"  4-NOV-1992 20:46:57.61
To:	MACRO32@WKUVX1.BITNET
CC:	
Subj:	Re: Undocumented SMG$ locator routine


In article <1992Nov4.193916.23882@magnus.acs.ohio-state.edu>, I wrote:

>
>[...]
>
>       CALL SMG$READ_LOCATOR ( KEYBOARD, ROW, COLUMN, KEYSTROKE )
>
>If the user hits a keyboard key instead of a mouse button, the corresponding
>terminator code is returned in KEYSTROKE and ROW & COLUMN are set to 0.  The
>only problem I've discovered is that if the user hits CTRL/Z, the EOF condition
>this produces is propagated to the next call to read input (e.g. SMG$READ_-
>KEYSTROKE, SMG$READ_COMPOSED_LINE, etc.).  I worked around this by special-
>casing CTRL/Z and calling SMG$READ_KEYSTROKE to "eat" the EOF condition.
>(SMG$CANCEL_INPUT was not sufficient.)
>
>[...]

        It's not that easy.  I discovered that on a DECterm, the extra call to
SMG$READ_KEYSTROKE works just fine.  On a standard text terminal (e.g. VT102),
the SMG$READ_KEYSTROKE should NOT be executed; on such terminals, SMG$READ_-
LOCATOR does not leave the CTRL/Z "waiting".  Until SMG$READ_LOCATOR is docu-
mented and/or fixed, I'm using the following (FORTRAN) code:

        LOGICAL MOUSE/ .TRUE./
        EXTERNAL SMG$_EOF
        SAVE MOUSE
        ...
        IF ( MOUSE ) THEN
            CALL SMG$READ_LOCATOR (keyboard-id, row, column, KEYCODE)
        ELSE
            CALL SMG$READ_KEYSTROKE (..., KEYCODE, ...)
        ENDIF
        IF ( KEYCODE .EQ. SMG$K_TRM_CTRLZ .AND. MOUSE ) THEN
            STATUS = SMG$READ_KEYSTROKE ( ...,KEYCODE, , 1, ...)
            IF ( .NOT. STATUS ) THEN
                MOUSE = STATUS .EQ. %LOC ( SMG$_EOF )
            ELSE
                MOUSE = .FALSE.
            ENDIF
        ENDIF
        ...

The jist of this code is that the first time through, it assumes a mouse is
connected.  If the user hits CTRL/Z in response to the call to SMG$READ_-
LOCATOR, SMG$READ_KEYSTROKE is called with a time-out value of 1 (second?).  If
the call times out or returns anything other than an end-of-file condition, it
asumes the user is on a terminal not equipped with a mouse; otherwise (SMG$-
READ_KEYSTROKE picked up an EOF w/o timing out), it continues believing it's
on a mouse-equipped terminal.  Since we don't have any real VT300-series
terminals, I can't test this code out on those platforms; if anyone feels so
inclined, I wouldn't mind hearing what happens when the above code is run on a
real VT3{3|4}0.  Also, just a reminder that we're running under 5.5-1.  If
anyone out there running 5.5-2 in field test could speak to whether the
SMG$READ_LOCATOR call is documented and/or fixed, I'd appreciate it.

                                                        Thanks,
                                                         Mike
--
 Internet: davism@KCGL1.eng.ohio-state.edu |
            -or- DAVISM+@osu.edu           |   These Thoughts, They Be Mine
   BITNET: DAVISM+@OHSTMAIL.BITNET         |