hanche@imf.unit.no (Harald Hanche-Olsen) (05/05/91)
Well, I finally got around to upgrading from TeX 2.991 to 3.0 on our
Apollos. The trip test failed dismally if I had optimising on, so I
turned it off and rexompiled, and the trip test was passed all right
(just as was the case with Version 2.991). So, optimizing still off,
I compiled the production version of TeX, and:
% initex plain '\dump'
This is TeX, C Version 3.0 (INITEX)
(/usr/local/lib/tex/inputs/plain.tex
Preloading the plain format: codes, registers, parameters, fonts,
more fonts, macros, math definitions,
Runaway definition?
->{\vbox {\ialign {$\m@th \scriptstyle ##$\crcr \not \mathrel
! TeX capacity exceeded, sorry [main memory size=3000].
********************* <--- What is this?
l.792 \not\mathrel{
\mkern14mu}\crcr
No pages of output.
Transcript written on plain.log.
Now, obviously, something is very wrong with that memory size!
tex/ctex.ch says @!mem_max=262140; and TeX thinks it has 3000?
Can any TeX source guru out there give me a hint? I don't even know
where to start looking for the source of this problem. Help!
The boring details:
Hardware: Apollo DN10000
Operating system: sr10.3 / bsd4.3
Compiler: cc v.6.8.p
- Harald Hanche-Olsen <hanche@imf.unit.no>
Division of Mathematical Sciences
The Norwegian Institute of Technology
N-7034 Trondheim, NORWAYhanche@imf.unit.no (Harald Hanche-Olsen) (05/06/91)
In article <HANCHE.91May5141039@hufsa.imf.unit.no> I wrote:
Well, I finally got around to upgrading from TeX 2.991 to 3.0 on our
Apollos. The trip test failed dismally if I had optimising on, so I
turned it off and rexompiled, [...]
% initex plain '\dump'
[...]
! TeX capacity exceeded, sorry [main memory size=3000].
As it turns out, this is what happens if you first make triptex, then
make "real" TeX. Apparently, making triptex changes tex/texd.h after
running the convert script and touching stamp-convert, and that change
will not be reversed when you build TeX. So TeX was built with
triptex parameters, which includes mem_top=3000.
I did 'make veryclean' and rebuilt, and things look rosy at the moment.
If course, make veryclean may be overkill, but then CPU time is
cheaper than my time ;-) Thanks to those who responded.
- Harald Hanche-Olsen <hanche@imf.unit.no>
Division of Mathematical Sciences
The Norwegian Institute of Technology
N-7034 Trondheim, NORWAY