[3184] in Athena Bugs
VAX eqn has a vertical spacing bug caused by a PCC bug.
daemon@ATHENA.MIT.EDU (Bill Cattey)
Wed Sep 13 00:38:55 1989
Date: Wed, 13 Sep 89 00:38:30 -0400 (EDT)
From: Bill Cattey <wdc@ATHENA.MIT.EDU>
To: bugs@ATHENA.MIT.EDU
Cc: geer@ATHENA.MIT.EDU
I've finally tracked down why VAX users of Andrew eq get ugly output
while RT users get pretty output.
It's not troff. The RT troff output when sent to the vax prints pretty.
It's not Andrew. Andrew generates the same eqn output on both the vax
and the RT.
It's not eqn. Although eqn generates different troff output on the VAX
than it does on the RT, the sources are the same (except for comments
saying that IBM owns the RT version.)
The version of eqn in at least VAX 6.3B generates incorrect results, but
when I compile one up using gcc -traditional, I get correct results.
----
There is code in eqn/shift.c which relies on particular precedence of
subtraction which seems to differ between PCC and (hc, and gcc).
Suggested actions:
1. We should, for ourselves, install a re-compiled eqn in /srvd/patch.
2. We should report to Berkeley that eqn gets mis-compiled by pcc on the
vax.
3. Some brave soul should put the parentheses in the right places in
eqn/shift.c test it, and send it back to Berkeley.
4. There are probably too many programs that rely on PCC's broken
precedence so PCC should not be changed.
-wdc
