Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!think!harvard!seismo!munnari!basser!ra!bdm
From: bdm@ra.anu.OZ (Brendan McKay)
Newsgroups: net.text
Subject: Re: Weird TeX behaviour
Message-ID: <195@ra.anu.OZ>
Date: Sat, 18-Jan-86 00:48:48 EST
Article-I.D.: ra.195
Posted: Sat Jan 18 00:48:48 1986
Date-Received: Tue, 21-Jan-86 01:15:13 EST
References: <777@lasspvax.UUCP>
Organization: Computer Science, Australian National Uni, Canberra
Lines: 38
Summary: reason for the problem noted

In article <777@lasspvax.UUCP>, garry@lasspvax.UUCP (Garry Wiegand) writes:
> 
>        A TeXnical question:    [ edited by bdm ]
> 
> \newskip \A
> 
> \A = 8pt \ifdim \A > 2pt \A = 2pt \fi \showthe \A	% answer is '8'!!!
> \A = 8pt \ifdim \A > 2pt \A = 2pt \fi \showthe \A	% answer is '2'!!!
>
 The explanation goes something like this:

When TeX has read the first "\A = 8pt" it doesn't know the assignment has
finished yet (it might continue  "plus 3pt" or something).  So it has to
expand the \ifdim ... \fi.  \A hasn't yet had it's value changed from the
initial 0pt, so the \ifdim ... \fi expands to nothing.  TeX now sees the
\showthe, realises the <dimen> has finished, assigns "8pt plus0pt minus0pt"
to \A and processes the \showthe.  It DOESN'T reevaluate the \ifdim ... \fi.
 
On the next line, the \ifdim ... \fi expands to "\A = 2pt " so TeX altogether
sees  "\A = 8pt \A = 2pt \showthe \A".

This behaviour may be counter-intuitive, but it's not a bug.  Equally weird
behaviour would be possible if the rules were changed to specify re-evaluation
of the \ifdim ... \fi.  Consider this one under both possibilitities:
	\newskip \X  \X = 8pt \ifdim \X > 2pt plus 1pt \fi \relax .

The solution is to insert \relax after the 8pt.


Brendan McKay.
Computer Science Dept., Australian National University, GPO Box 4, Canberra,
ACT 2601, Australia.

ACSnet: bdm@anucsd.anu.oz	  		ARPA: bdm%anucsd.anu.oz@seismo	
CSNET: bdm@anucsd.anu.oz@csnet-relay.csnet	JANET: anucsd.anu.oz!bdm@ukc
UUCP: {decvax,vax135,pesnta,eagle}!mulga!anucsd.anu.oz!bdm
      {seismo,ubc-vision,ukc,mcvax,prlb2}!munnari!anucsd.anu.oz!bdm
	[ UUCP routes through munnari prefered ]