earle@smeagol.UUCP (Greg_Earle) (10/10/85)
I have a question for the Sun gurus. When inside the window system,
executing "tset -QS" directly from the shell gives output that
'knows' about the size (lines, columns) of the current window:
e.g.
% tset -QS; echo
sun Mu|sun:li#50:co#76:cl=^L:cm=\E[%i%d;%dH:nd=\E[C:up=\E[A:am:bs:km:mi:ms:
(etc.)
rs=\E[s: %
(line broken up for readability). Note the entries for li and co.
But lo and behold; if you should try to do anything with this output, like
put it between `` for use with eval, or even through a pipe or a redirected
output, the output suddenly mysteriously changes back to a 'normal' sun
termcap entry:
% tset -QS | cat
sun Mu|sun:li#34:co#80:cl=^L:cm=\E[%i%d;%dH:nd=\E[C:up=\E[A:am:bs:km:mi:ms:
( etc.)
rs=\E[s: %
Or:
% tset -QS > /tmp/foo
% cat /tmp/foo
sun Mu|sun:li#34:co#80:cl=^L:cm=\E[%i%d;%dH:nd=\E[C:up=\E[A:am:bs:km:mi:ms:
( etc. )
rs=\E[s : %
How/Why is this happening? more(1) seems to know about the length of the tty
subwindow; you get everything from the top of the window down to the last
line before more stops. Does it get it directly from the window subsystem?
One reason I ask is that it seems like it could be a wonderful feature;
you could put a series of statements in your .cshrc file (like the
example given in the man page for tset with option -S) so each program
that used $TERMCAP would know about the window limits if it wanted
(I was thinking of modifying vnews(1) to do this, instead of the fixed
# of lines you get).
For a minute, I thought the old "can't set your parent's env from yerself"
but that can't be it; you can muck all your other environment variables to
your hearts content inside the (non-login) shell windows.
Thanks in advance.
Greg Earle
...!{{decvax,ucbvax}!sdcsvax,hplabs,allegra,trwrb}!sdcrdcf!smeagol!lorien!earle
...!{ihnp4,decvax,ucbvax}!sun!tsunami!smeagol!lorien!earlechris@umcp-cs.UUCP (Chris Torek) (10/13/85)
The answer no doubt lies in the fact that the window size is obtained via an ioctl: `TIOCGSIZE'. Presumably Sun's tset is performing this on its standard output, and inside backquotes this is tied to a pipe rather than the terminal. If you have source, you may be able to fix tset by making the ioctl happen on 0 or 2 (stdin or stderr). Sun changed their termcap routines to know about the `li#' and `co#' entries, and to return the appropriate values based on the window size, if available. Thus programs like `more' work without modification. -- In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 4251) UUCP: seismo!umcp-cs!chris CSNet: chris@umcp-cs ARPA: chris@mimsy.umd.edu
shannon@sun.uucp (Bill Shannon) (10/14/85)
Within the termcap library, whenever you do a tgetent, it tries to find out the size of the current window. It does this by doing an ioctl on stdout. Almost always, when using termcap, stdout is the tty/window you're going to do termcap things to. If the ioctl succeeds, tgetent modifies the normal termcap entry to insert new "li" and "co" entries at the front (where they'll be found before the attributes in the normal termcap entry). This explains why it fails when you pipe the output through cat but succeeds in almost all real cases. Bill Shannon P.S. Not mentioned above is the fact that if the window size changes after doing the tgetent, the program must've set itself up to notice the change (catch the SIGWINCH signal) and do the ioctl itself to get the new window size.
peter@graffiti.UUCP (Peter da Silva) (10/15/85)
> I have a question for the Sun gurus. When inside the window system, > executing "tset -QS" directly from the shell gives output that > 'knows' about the size (lines, columns) of the current window: > But lo and behold; if you should try to do anything with this output, like > put it between `` for use with eval, or even through a pipe or a redirected > output, the output suddenly mysteriously changes back to a 'normal' sun > termcap entry: Sounds like tset is using stdout for the ioctl that reads the real screen size. OOPS (as tgoto(3) said after I did something similar).