Re: about the tests

[Karl Berry]

    I believe there are some missing, like with --no-headers, for docbook, 
    maybe pdf, but I am not knowledgable enough to know what is worth testing.

makeinfo --no-headers is worth checking, that's the plain text output.
  (There are warnings with texinfo.txi though, sigh.)
makeinfo --no-headers --html is not worth checking, IMHO.
makeinfo --docbook, let's not worry about that for now.  I assume it'll
  be easy enough to add later.
texi2dvi --pdf is worth checking.

Don't bother running dvips, its output will depend on the exact details
of the person's TeX setup.  We don't want tests to fail based on that.
The DVI output is all that matters.

As for the timestamp in the DVI file, it is always in the same place in
the file.  DVI files start with (see dvitype.web et al.):
247 (decimal) |i[1]| |num[4]| |den[4]| |mag[4]| |k[1]| |x[k]|.
So, skip 14 bytes (247 I num den mag), read the next byte to get the
length K, and zero out the next K bytes.

No, there's an easier way.  In the dvitype output, the comment is output
starting with an apostrophe.  grep -v \^\' and that should be it.
Also skip the first line, which gives the dvitype version number.

I guess we'll see if there are other problems.

If you could make the dvi and/or pdf checking separate scripts, they
would be useful in other contexts as well (namely TeX distributions).

Thanks!