[texhax] mktexpk fights cmcsc, ibycus4-greek ... indeed

[Uwe Lück]

At 16:11 25.09.05, Barbara Beeton wrote:
>...
>
>the reason i'm responding to your message is that
>if there is really a problem with generating cmcsc17,
>and you really do get the "this can't happen" error
>with that font, then i'd like to explore this further
>to determine if it really is a metafont bug that
>should be reported to knuth.  can't help you with
>wasy2 or ibycus-greek (nor can i; sorry), but cmcsc17
>should be close enough to the knuthian fonts that he
>will look at the report (in 2007).
>
>it's also possible the problem is in the local
>environment adaptation, in which case it may or may
>not be appropriate to send it to knuth; but that
>can be determined by trying it out on different
>systems, and i can help find people to do that.
>
>please let me know if you can reproduce the problem
>with cmcsc17, and if so, write up a report on just
>that test, zip up the files that were used along with
>the log and send them to me.  i'll also watch texhax
>to see what turns up there.

Barbara and Karl,

you may have seen here that the problem turned out to be
a MiKTeX feature in the first instance. Incidentally, however,
it led me to a problem with mktexpk. I have just learnt from
the TeXpert at the computing center that the mktexpk I have
used is just from the current TeXLive distribution, so will you
be please care for the latter if there indeed is a bug?

Now I did:

mktexpk --dpi 600 --bdpi 600 --mag 0 --mfmode ljfour fibr84

The screen then was crowded with error messages,
it seems that division by 0 is tried some hundreds of times,
I am sending the .log containing 15 200 lines separately.
fibr84.tfm and fibr84.gf are announced, but fibr84.gf is _not_
there.

Yet

mf '\mode=ljfour; mag=magstep(0);' input fibr84
gftopk fibr84.600

works (I just tested the .pk with my 300dpi printer, and it
seems to be OK). So I think mf is OK, there is no bug to
be reported to Knuth, and I won't win anything from him
(as the METAFONT book somewhere announces).

Now, I have tried mktexpk on cmcsc10 at magsteps 2
and 5 like above, and both failed! Unfortunately, I cannot
get the .log's, and "more" didn't help, I only see that
there are no errors in processing the characters, yet the
output is cmcsc10.1200gf, e.g., a proof output, i.e.
The final message is that it failed indeed.

I cannot go on this way because the outcomes seem
to go to directories at the computing center to which
I don't have access -- well, I have another idea, but
now I must go shopping.

Best,
Uwe.