[tex-live] dvips for windows behaviour

Karl Berry karl at freefriends.org
Thu Nov 29 00:18:24 CET 2007

    +       if (cmdsave == 129 || cmdsave == 134) {

The parsing in skippage.c seems more generally flawed.  It was just
accidental that the problem wasn't exposed on Unix, because (I guess) we
enabled Omega by default.  (By the way, this has nothing to do with the
bug, but why not enable Omega parsing on Windows, too?  Does it conflict
with Japanese usage?)

Of course sample2e.dvi does not actually contain any "Omega" opcodes, as
the dvitype output shows.  (Actually these opcodes are perfectly valid
DVI, even if TeX doesn't use them.  Anyway ...)

The problem, it seems to me (just looking at the code, I didn't do any
debugging) is that the switch in skippage.c falls through to the Omega
test more or less all the time, hence the complaint about the random 121
byte.  This just seems wrong.

So I think the best thing to do is just remove the noomega test
entirely, and let 129 and 134 be handled just as the other (144, ...,
236) two-byte cases.  We're skipping bytes here.  The time to complain
about an unhandled opcode is when the opcode is actually (attempted to
be) processed.

Does that work?  Wdyt?


More information about the tex-live mailing list