Skip to content

Could not create DVI version

7 messages · Brian Ripley, Paul Gilbert, Friedrich.Leisch@univie.ac.at

#
In Solaris with R-1.7.0beta from 2003-04-01 I still have a problem with
"R CMD check" after I build a package:

* creating setRNG-manual.tex ... OK
* checking setRNG-manual.tex ... ERROR
Could not create DVI version.
This typically indicates Rd problems.

I can manually gererate the dvi from setRNG-manual.tex, so I don't think
my latex is defective as I speculated last week. The file texput.log
indicates

This is TeX, Version 3.14159 (Web2C 7.3.1) (format=latex 2001.1.12)  1
APR 2003 12:09
**\nonstopmode\inputsetRNG-manual.tex
! Emergency stop.
<*> \nonstopmode\inputsetRNG-manual.tex
                                       
*** (job aborted, no legal \end found)


I don't know what this is suppose to mean, or whether this is where I am
suppose to look for the source of the problem. Other than the ongoing
occassional segfault problems with sweave (gc?), everything works fine
in Mandrake 9.0 and 9.1 (with R-1.7.0beta from 2003-03-31).

Paul Gilbert
#
On Tue, 1 Apr 2003, Paul Gilbert wrote:

            
I've seen that several times: I think there is an Sweave-related gc 
problem.
#
I can run R CMD check with current CVS version of R on dse_2003.3-1 on
Solaris 8 successfully (using gcc 3.2, but that's not relevant here).

There is

  est.wt.variables                  text    html    latex
     missing link(s):  est.min.sqerror

but that's a long-standing error in your packages.


This does look like a problem in your Solaris setup.
On Tue, 1 Apr 2003, Prof Brian Ripley wrote:

            
Your Perl?
^ 
I get

**\nonstopmode{}\input{setRNG-manual.tex}

so something is wrong.
#
Prof Brian Ripley wrote:
I just confirmed that I get the same problem with that version of dse.
I'm using  SunOS 5.8 and perl, v5.6.1 built for sun4-solaris.
Fixed in my sources now.
Yes. Any suggestions of something else I should try?

Thanks,
Paul
#
On Tue, 1 Apr 2003, Paul Gilbert wrote:

            
What locale are you using?  Might it be worth using 

setenv LANG=C  (or your shell's equivalent)
R CMD check dse

?  I'm wondering about the input encoding to LaTeX: not that I know much 
about it as an Anglophone.
#
Brian

I tried LANG=C without success, but it seems there is something in our latex
startup script that causes the problem. The latex call actually executes a csh
script that sets local variables then executes the latex binary. If I bypass the
script and make a direct call to the binary then the problem goes away.

Thanks for your help.

Paul
Prof Brian Ripley wrote:
#
[...]

  > I've seen that several times: I think there is an Sweave-related gc 
  > problem.

Anything reproducible?

Paul nd I tried to nail down the problem several times ... but as the
bug didn't shon on our machines, it was impossible for me to track
down the bug yet.

As Sweave uses only S code, it can only be an R bug triggered by
Sweave ...

.f