Ulrich Kohler ([email protected]) wrote about some difficulties he
had installing Stata 10 on various distributions of Linux.
All of Ulich's difficulties can occur only under Linux/Unix and can be
usually avoided by installing the statically-linked version of Stata
from the CD.
Here are the specifics.
Ulrich begins:
> I have now installed Stata 10 for Linux on 2 (3?) different Linux
> systems. Just thought to share some of my experiences:
>
> (1) Stata 10 on Xubuntu, "Dapper Drake"
The first thing Ulrich mentioned was that he could not install
the software directly from the CD on his first system, and he
mentioned that the advice in the Getting Started manual's
troubleshooting section solved the problem.
To elaborate, some Linux systems will not allow scripts to be
executed from a CD. They disable 'execute' permission for all
files on the CD. On such Linux systems, it is impossible to
install Stata (10, 9, 8, it does not matter) directly from the CD.
On such systems, the Getting Started manual recommends that the
installation files from the CD be first copied to a location on
the hard drive before beginning the installation. Ulrich did
this, and the installation proceeded as expected.
Ulrich then wrote about some library issues:
> - After installation console Stata works, but xstata didn't start due to
> a missing libary, namely "libtiff.so.3". ...
Ulrich then went on to describe some truly heroic machinations that he
performed to install new operating system libraries. After which, his
installation worked.
While I admire Uli's resourcefulness and tenacity, I would not
recommend that others follow his approach. Instead, simply install the
statically-linked version of Stata from the CD rather than the
dynamically-linked version. This approach is recommended by both the
installation program and in the Getting Started manual.
Here are the details.
The problems Uli encountered with dynamically-linked version is common
on Linux systems with Stata and other software that uses shared
libraries. Various Linux systems often have different versions of
libraries on them which can make it difficult for an application to
run.
Because of this, the Stata 10 installation CD includes both
dynamically linked and statically linked versions of Stata for Linux.
The dynamically linked version attempts to use the shared libraries
on a user's system. The statically linked version includes some
important libraries within itself and so does not rely on all of the
libraries being on the user's system. Linux systems are so widely
known for such library issues that the Stata 10 installation program
recommends
"If you encounter any problems with the dynamically linked
version, reinstall using the statically linked version."
If Ulrich had contacted Stata Technical Support, their first
advice to him would have been to uninstall the dynamic version
and reinstall using the static version. If he still encounted
problems, they would then walk through some steps with him to
either find or install the necessary libraries on his system.
Ulrich wrote that he had an issue with the font appearing on his
systems:
> (4) Finally, there are two issues with Xstata on the two systems:
>
> (a): Stata used a non-monospaced font for results and help files,
> although the font mentioned in the preferences menue was "fixed 10".
> However, when I clicked on the font, the highlighted font in the list
> was "sans 10".
We have not heard of this from any other users nor can we reproduce
it on any of our own Linux systems. It is possible that the default
font Stata tried to use was not available on his system, and so the
system defaulted to what it thought was the closest match. Ulrich
should contact Stata Technical support at [email protected]
so that they can help diagnose what is unique about his system.
As a workaround, Ulrich should be able to change the font in
his Preferences to one of his liking.
Ulrich then wrote about an error message he received:
> (b):
>
> . sysuse auto
> . reg weight
> . reg
> 95,0 found where number expected
> r(198);
>
> . logit foreign
> . logit
> 95,0 found where number expected
> r(198);
>
>
> Is this a bug? I noticed that the syntax diagram of -regess- is
> -regress depvar [indepvars]-, which somhow implies that -regress-
> without varname is no longer valid syntax. However, the help-file also
> mentioned
We have heard no other reports of this and cannot reproduce it.
What Ulrich is trying should work fine and does work fine on
a fresh Stata installation just tested on a Linux machine.
My first thought was that Ulrich had -set dp comma- to use a comma
as the decimal point in Stata and that something was going wrong
as a result. However, setting -set dp comma- and typing the
commands Ulrich used results in no problem.
Ulrich should contact Stata Technical Support at [email protected]
with the above issue. They will help him diagnose it, and we will
report back to the list with the cause and solution.
--Alan
([email protected])
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
