Quits without failing with "Unknown graphics extension: .eps"

[GoogleCodeExporter]

I'm seeing some really weird behavior for a file that hasn't been fixed to 
remove ".eps".  Note that there isn't the error about an unknown extension, 
and there isn't a "Success" line being printed:

amcnabb@prodigy:+mjg82-papers/byu_studies% make clean
amcnabb@prodigy:+mjg82-papers/byu_studies% make poster.pdf
NOTE: You may ignore warnings about the following files:

     poster.d

Makefile:2051: poster.d: No such file or directory
= poster.tex --> poster.d poster.pdf (1) =
Missing argument. Try `kpsewhich --help' for more information.
make: `poster.pdf' is up to date.
amcnabb@prodigy:+mjg82-papers/byu_studies%

Even more oddly, one of the ten times I ran it, it gave the correct error 
messages, for example:
--- Graphics extension error:
     ./poster.tex:252: LaTeX Error: Unknown graphics extension: .eps.
--- If you specified the extension explicitly in your .tex file, try 
removing it.

When I run pdflatex by hand, it seems to fail just the way it should:

...
! LaTeX Error: Unknown graphics extension: .eps.

See the LaTeX manual or LaTeX Companion for explanation.
Type  H <return>  for immediate help.
 ...                                              

l.118   }

?

I'll attach the log file in case it's helpful.

Original issue reported on code.google.com by amcna...@gmail.com on 11 Mar 2010 at 9:54

Attachments:

• poster.log

[GoogleCodeExporter]

Changing to "New", since I haven't actually looked at it yet.

Original comment by shiblon on 12 Mar 2010 at 5:23

• Changed state: New

[GoogleCodeExporter]

Is this still happening, and is there a minimal example that you can post (a 
.tex file and 
.gpi or something) that reproduces this behavior?

Original comment by shiblon on 12 Mar 2010 at 7:43

[GoogleCodeExporter]

I just did an "hg pull -u", and I'm still seeing the problem.  I'll point you 
at an 
example by private email.

Original comment by amcna...@gmail.com on 12 Mar 2010 at 11:07

[GoogleCodeExporter]

The best thing would be for me to see the .log output when you run the 
following 
command:

pdflatex -interaction=batchmode poster

Then I can look at it and hopefully see what's missing.

Original comment by shiblon on 15 Mar 2010 at 4:04

[GoogleCodeExporter]

Here is the new log file.

Original comment by amcna...@gmail.com on 15 Mar 2010 at 4:15

Attachments:

• poster.log

[GoogleCodeExporter]

Weird!  That should totally work!  The very first thing that happens when 
building the 
.pdf target is to colorize the latex errors.  If anything is output (because it 
found a 
bad graphics extension, for example), then it fails with a fatal error.

Odd stuff.

So, here's the next thing that would help.  Run 

make clean; make -d poster &> make.out

And send me that :)

Original comment by shiblon on 15 Mar 2010 at 5:33

[GoogleCodeExporter]

Attaching make.out as requested.  If things get too crazy, feel free to use the 
copy 
on guru that I emailed you about.

Original comment by amcna...@gmail.com on 15 Mar 2010 at 5:47

Attachments:

• make.out

[GoogleCodeExporter]

The make.out file looks like the result of calling make SHELL_DEBUG=1.  What I 
really 
need is the output of make -d.

Original comment by shiblon on 16 Mar 2010 at 1:33

[GoogleCodeExporter]

Scratch that - I was looking at the wrong file.

Original comment by shiblon on 16 Mar 2010 at 1:34

[GoogleCodeExporter]

Fixed the problem.  Very, very subtle interaction between make functions (e.g., 
$(wildcard)) and shell commands in command execution.

See the diff for r0eb58b650c83

This and the fix before it should make your experience much more stable.

In fact, I found an issue that may make it possible to add __gray back in.  
Let's discuss.

Original comment by shiblon on 16 Mar 2010 at 2:03

• Changed state: Fixed

[GoogleCodeExporter]

I now have a beautiful error message. :)

Original comment by amcna...@gmail.com on 16 Mar 2010 at 4:10

[GoogleCodeExporter]

Original comment by shiblon on 16 Mar 2010 at 4:19

• Changed state: Verified