Muted "LaTeX Error" when the class file is missing

[shiblon]

Originally reported on Google Code with ID 98

When a class file is missing, pdflatex reports the error as follows:

! LaTeX Error: File `IEEEtran.cls' not found.

However, this error is not being passed on by the Makefile, which instead reports:

% make
NOTE: You may ignore warnings about the following files:

     paper.d

Makefile:2265: paper.d: No such file or directory
= paper.tex --> paper.d paper.pdf.1st.make (0-1) =
make: *** [paper.d] Error 1
%

If it would be helpful, I could produce a minimal test case, but it's easy enough to
make one by taking any latex file and replacing the document class with gibberish.

Reported by amcnabb8 on 2010-11-04 22:00:17

[shiblon]

By the way, I think the Makefile might be catching the "! LaTeX Error: File" and trying
to find if there's a way to autogenerate the file, but when it realizes it can't do
this, it just says "Error 1".  That's just an idea--I don't know if it's helpful.

Reported by amcnabb8 on 2010-11-04 22:02:25

[shiblon]

Yes, that's exactly what's happening.  What is confusing me is that you aren't getting
a "no rule to build IEEEtran.cls" error, but an Error 1 instead.

Make really is supposed to report these better than it is.  I wonder what is going
on.

I'll see if I can make a minimal test file to trigger this.

Reported by shiblon on 2010-11-05 19:40:54

• Status changed: Accepted

[shiblon]

Try rcf3fcf70a99e and see if that floats your boat.

Reported by shiblon on 2010-11-05 20:21:19

• Status changed: Started

[shiblon]

That seemed to do it.  By the way, will the fix work for missing graphics files, too?
 I think I've occasionally run into the same error when there's a missing graphics
file that can't be autocreated.

Anyway, for the .cls case, I can confirm that the problem is fixed.

Reported by amcnabb8 on 2010-11-05 20:31:48

[shiblon]

I'm not sure whether this will affect graphics that can't be built or not.  It seems
like it *shouldn't*, but there is enough complexity here, and I'm far enough removed
from using it, that I can't say for sure.

Speaking of which, this project needs another maintainer.  Volunteers?

Reported by shiblon on 2010-11-08 17:35:45

• Status changed: Verified

[shiblon]

If it were in Python, I just might volunteer, but make makes my sed head hurt.  I mean
it makes my sad head hurt.  I think.

Reported by amcnabb8 on 2010-11-08 19:58:32

[shiblon]

I'm about >< this close to starting over in Python.  If only I had time to do that instead
of fix bugs ;-)

I jest, of course.  Bugs are important, and I am going to fix them (someday).  But,
I would *love* to brainstorm with you about how a Python version would look, especially
since you are deep into using it right now.  If you have some time for a chat sometime,
let's start that project.  It would be fun.

Meanwhile, I'm going to leave this bug closed.  If you find it bites you again for
other missing files, let's reopen it.

Reported by shiblon on 2010-11-11 15:14:26

[shiblon]

I think that doing it in Python sounds fun.  I'm not sure where the dependency resolution
stuff would come from (it would be nice if there were a well-maintained existing library
that did this), and it would be nice to see what should or shouldn't be borrowed from
Rubber.

Reported by amcnabb8 on 2010-11-11 17:47:09

[shiblon]

Cool.   Let's take this discussion out of the tracker, then, and see what we can come
up with.

Reported by shiblon on 2010-11-11 18:13:37

[shiblon]

I just had cause to use this and found that a *ton* of errors are passing silently,
now.  As a result, I'm reopening this bug and have just finished a major code restructuring
project to facilitate finding these issues.

There is no longer a "Makefile" in the source code, only a Makefile.in and a whole
bunch of supporting bits and pieces.  Right now there are only a few things factored
out, and most of them are .sed files.  To get the Makefile, just run "./build".

Now, to debug things like latex error output, you can just do

./run_sed.py colorize-latex-errors.sed yourtexfile.log

And it will show you what the sed script is doing.  You'll also notice that the sed
script is actually a valid script, so you could *almost* just run it directly (it might
not look quite right, but it will still give you a good idea of what's going on).

Anyway, take a look at the output of that with a broken log file.  It should help us
to zero in on problems like these.

Reported by shiblon on 2010-11-22 18:21:39

• Status changed: Started

[shiblon]

By the way, I just pushed another tip.  I think this one works better.  Just update,
./build and give it a whirl.

In other excellent news, colorize-latex-errors.sed is now heavily, heavily commented,
and should therefore be more accessible to mortals.  Doing this build step from source
to create the makefile allows me to do that because it makes post-processing comments
away possible.  Nifty.  Anyway, have a look, give run_sed.py a whirl, and let me know
what errors you aren't finding in your log.

Reported by shiblon on 2010-11-23 01:03:09

[shiblon]

Just to follow up with this, I believe that this issue is fully put to bed with the
completion of issue 113.  As for redoing in Python, that still sounds fun, but at least
we're making progress on things as they are.

Reported by shiblon on 2011-01-04 15:40:11

• Status changed: Fixed