Corrupted NFSS tables' is written to stdout, but document is correctly produced, and the whole process is normal (i.e. no error) with Kile

[shiblon]

Originally reported on Google Code with ID 84

Hi,

When compiling the files at
`http://gitorious.org/angstrom-manual/angstrom-manual/trees/master' with
the most recent latex makefile, I receive this at stdout:

==
! Corrupted NFSS tables.
wrong@fontshape ...message {Corrupted NFSS tables}
                                                  error@fontshape else let f...
l.1 \AA
        ngstr\"{o}m was started by a small group of people who worked on the...
This error message was generated by an \errmessage
command, so I can't give any explicit help.
Pretend that you're Hercule Poirot: Examine all clues,
and deduce the truth by order and method.

! LaTeX Error: Missing \begin{document}.

! Emergency stop.
<*> history.tex

*** (job aborted, no legal \end found)
==

Strange, as the resulting PDF is correctly produced. Furthermore, there
cannot be any errors in the source files, as the whole process goes
normally under Kile. Why? Files are available at the git repo. (The
included makefile is not the most recent one, but I used the up-to-date
makefile to try compilation too.)

Reported by merciadriluca on 2010-05-14 16:35:24

[shiblon]

Kile doesn't output all of the errors that it should.  I have had other bugs submitted
where Kile was silently 
pretending that errors didn't happen, and people thought everything was working fine.
 In reality, there were 
subtle problems with their documents, and the makefile properly caught them where Kile
did not.

When building with Kile, take a look at your .log files and I bet you'll see the errors
in there.

Reported by shiblon on 2010-05-17 17:34:12

[shiblon]

Sorry, I should have mentioned it, but I had checked kile's compilation's resulting
log after compilation. You'll find it attached. 

Reported by merciadriluca on 2010-05-17 17:52:48

---

• Attachment: angstrom-manual.log

[shiblon]

Note also that the last resulting .log file for the project, but, this time, using
the makefile, gives no error. 

Reported by merciadriluca on 2010-05-17 17:54:07

---

• Attachment: angstrom-manual.3.log

[shiblon]

Interesting.  Are you still having this problem, or did it somehow resolve itself? 
I don't really know what an NFSS table is...

Reported by shiblon on 2010-06-17 20:39:59

[shiblon]

I don't know neither what it is. I'll ask on comp.text.tex. But I am still having the
issue, yes.

Reported by merciadriluca on 2010-06-18 11:14:38

[shiblon]

Well, at http://tug.org/pipermail/yandytex/2004-July/000902.html some things are explained,
and at http://www.latex-project.org/guides/lc2fr-apb.pdf, p. 10/61, it is explained
that `Corrupted NFSS tables' actually means that LaTeX has tried some font substitutions,
and has detected some inconsistency in its intern tables. This error habitually comes
when a font substitution has been launched and that substitution's rules contain some
loop (sub-circular declarations), or when the arguments of the default substitution
(in the current coding) point to a group of inexisting fonts.

It is certainly due to the LaTeX makefile, because my related .log file contains no
related errors/warnings, after Kile's compilation's invocation.

Reported by merciadriluca on 2010-06-18 11:29:58

[shiblon]

This last bit might be overstated.  Have you tried just running latex directly on your
file and then checking the log?  The log file is overwritten with subsequent invocations,
so it probably only shows up (and gets swallowed) in the first invocation that Kile
performs.

Reported by shiblon on 2010-06-18 12:52:29

[shiblon]

Yes, and it shows nothing. :(

Reported by merciadriluca on 2010-06-18 13:02:54

[shiblon]

OK - I'm happy to look at this in greater detail, but unfortunately the log entry isn't
giving me the information that I need.  If it is possible, can you produce and attach
a minimal document that has this behavior?  Then I can try it on my end and attempt
to track it down.

Reported by shiblon on 2010-06-18 13:07:14

[shiblon]

Well, I would like to, but I cannot reproduce this problem. This problem arises with
a very long preamble, and a very long (> 1000 pg.) document containing lot of equations.

Reported by merciadriluca on 2010-06-18 13:59:37

[shiblon]

:-/

So, let's see.  Would demoting this to a warning satisfy your needs for now?  That
way it will at least display, but won't keep your document from compiling.

Reported by shiblon on 2010-06-23 19:25:30

[shiblon]

Nice idea. :)

Reported by merciadriluca on 2010-06-23 19:52:03

[shiblon]

Demoting this to a warning won't work, since LaTeX itself is barfing on the files. 
I'll have to find something else to do.  I'm currently downloading a tarball of the
project to see what I can find.

Meanwhile, some questions that might generate leads:

1) Is Kile using the same "pdflatex" binary as the makefile?  Or is it using an alternative
installation not in the $PATH?

2) Does this happen every time, or intermittently?

3) If you fully clean things up before building, does it change the behavior?

4) The makefile is really not happy with subdirectories.  Sometimes things work okay,
but generally speaking, having source files in subdirectories is fraught with peril.
 Have you tried flattening your directory structure?  That might tell us something,
too.

Reported by shiblon on 2010-06-29 19:35:22

[shiblon]

Also, I noticed that you are using 2.1.34, but 2.2.0-rc2 is a much more recent (and
in many ways completely reworked) version of the Makefile.   Have you tried that instead?

Reported by shiblon on 2010-06-29 19:36:49

[shiblon]

1) Kile uses the same PDFLaTeX, but I don't understand why you speak about PDFLaTeX,
as it is here question of the latex -> dvi2ps -> ps2pdf route. If you wanted to speak
about these three tools, yes, the sames are shared for Kile and the makefile.

2) Every time.

3) No. I tried.

4) I know. But it worked like a charm some days before the bug I encountered (happened);
I don't know what I changed since then, but that was not that much (and not to the
preamble). So, this must not be the directories.

5) I'll try it now.

Reported by merciadriluca on 2010-06-29 19:41:39

[shiblon]

5) Well, with the most recent Makefile, things do not get better:

==
--- Graphics extension error:
     ! LaTeX Error: Unknown graphics extension: .eps.
--- If you specified the extension explicitly in your .tex file, try removing it.

--- Graphics extension error:
     ! LaTeX Error: Unknown graphics extension: .eps.
--- If you specified the extension explicitly in your .tex file, try removing it.

--- Graphics extension error:
     ! LaTeX Error: Unknown graphics extension: .eps.
--- If you specified the extension explicitly in your .tex file, try removing it.
==

while Kile compiles the whole without any problem (and without such comments in the
.log). Sorry, I'd have preferred it to be okay. :(

Reported by merciadriluca on 2010-06-29 19:46:02

[shiblon]

First, I was actually using 2.1.40 as an `old makefile.'

Second, fresh news. Here you are:

* With 2.1.40, everything now compiles like a charm (don't know why): no error as before,
* With 2.2.0-rc2, I get the error of my Comment #16.

Reported by merciadriluca on 2010-06-29 19:52:37

[shiblon]

The error message using the latest rc2 of the makefile actually contains the solution
within it.  Instead of specifying, e.g.,

{_img/kexecboot-0-4-1.eps}

instead, take the extension away and specify it like this:

{_img/kexecboot-0-4-1}

I'm happy that the later version of the old makefile is working for you, and I'm confident
that the new rc2 version will be good, too.  You'll notice that cleaning doesn't work
properly for images in subdirectories - not much I can do about that, unfortunately,
but otherwise you should be fine.

I'm marking this WontFix, but do feel free to reopen it if you still have problems.

Reported by shiblon on 2010-06-29 20:00:20

• Status changed: WontFix

[shiblon]

Thanks. But I don't agree for the eps problem. I know that .eps is the default extension,
and that, once you define a default extension for images, appending other extensions
(even if these are the same) to the images' calls might cause trouble. But, here, I
never declared .eps as my default extension. So, why do I get these messages? The Makefile
might consider .eps as the default extension, but that obliges (since the latest rc2)
the user not to append .eps extensions to images' calls, which is not consistent with
the fact that it was not necessary in previous releases.

Reported by merciadriluca on 2010-06-29 20:06:21

[shiblon]

The move to pdflatex has made this necessary: there's no way to get it to do the right
thing (from a dependency detection point of view) with specified extensions using pdflatex
vs. plain old latex.  Therefore, the removal of the extension has to stay.  The good
news is that not specifying an extension still works in both worlds.

In greater detail, the problem is this: if you specify an extension in your file, then
pdflatex *quits immediately* if that file is not found, even in batch mode!  Obviously,
this breaks all kinds of auto-generation stuff.  If you don't specify the extension,
then pdflatex emits an error in the log and moves on, which is what the makefile uses
to determine which graphics need to be built and included.  Removing the extension
is the only way to make it do the right thing.  I even tried monkeypatching latex to
get around this limitation, to no avail.

Sorry I can't do much about that - I know it sort of feels ugly, but we're running
headlong into LaTeX problems here, and I'm very much constrained by the underlying
toolset.

Reported by shiblon on 2010-06-30 13:33:22

[shiblon]

No problem. :-)

Reported by merciadriluca on 2010-06-30 15:23:58