beede@hubcap.UUCP (03/09/87)
While trying to get TeX running on a Celerity UXIX system, I
encountered some difficulty reading binary files. Apparently
Celerity Pascal is _almost_ identical to Berkley Pascal, down
to sharing the same structure for file descriptors. However,
the Berkley version wouldn't work when reading binary files.
To ensure portability between machines, TeX handles binary files
a byte at a time (thus ensuring that there is no big/little-endian
conflict between machines. The Berkley version declared a file
as ``packed array of -128..127,'' then reads it using assignments
from the file pointer, in a loop (simplified) like
while not eof(f) do
begin
array[i] := f^;
get(f); i := i+1
end
Some values were lost, and I had no luck fixing the matter, so I
wrote an external C routine to read a byte and return it. I changed
the (simplified) loop to
while not eof(f) do
begin
array[i] := newread(f);
i := i+1
end
which STILL lost some values (always 255's--these are really supposed
to be unsigned bytes. The old version tested and added 256 on negative
values).
The final problem proved to be that eof(f), when the next byte was
255, returned false after MOVING THE FILE POINTER. i.e., one byte
was lost.
My question: is this expected behavior, or am I correct in assuming
that someone cheaped out on the code generation and used some character-
oriented routine to read from anything that looked like a file of bytes?
Thanks.
--
Mike Beede UUCP: . . . ! gatech!hubcap!beede
Computer Science Dept.
Clemson University
Clemson SC 29631-1906