On Tue, May 23, 2006 at 01:56:15PM +0200, Jean-Marc Lasgouttes wrote:

> >>>>> "Enrico" == Enrico Forestieri <[EMAIL PROTECTED]> writes:
> 
> Enrico> I found that it will be recognized if you have a \path_prefix
> Enrico> entry in lyxrc.dist.
> 
> Enrico> This fact hints at a wrong way of dealing with \path_prefix,
> Enrico> as user changes will be recorded in preferences and not in
> Enrico> lyxrc.dist.
> 
> Originally, path_prefix was intended as something for the installers.
> In theory, when installers are perfect, there is no need to change it
> in prefs. 
> 
> I tend to think that people who change path prefix could as well
> change their environment, but I might be missing reasonable uses. Actually
> I'd even propose to remove path_prefix from the prefs UI.

I think that path_prefix is quite useful on Windows as it is a mean
to overcome the paths with spaces problem. Currently you can't set
a viewer in preferences if its path contains spaces. So you simply
add its path to path_prefix and specify the viewer using its name only.

This actually works if the viewer is not to be found by configure.py.

IMO, it is an error that the path_prefix set in preferences (as opposed
to lyxrc.dist) is not used by configure.py. I was bitten by this when
I tried to test the native version of pplatex.

However, things seems to be a bit more complicated than this, as I am
not sure that path_prefix is not actually used when you configure from
inside LyX (as path_prefix should already be prepended to PATH).

I am sure that it is not used when configure.py is launched by LyX
when it is starting. Actually I am unable to test if it is really
used when configuring from inside LyX because configure.py hangs for
me in this case. I think it is due to the fact that I mostly use
cygwin programs (I only have miktex and imagemagick as native apps).

I know that the problem is a stalled latex process and I think that
it is due to the port of configure to python. In the bourne shell
version a redirection of the standard input from "/dev/null" was
surely helping in avoding the problem whereas the -interaction=nonstopmode
option doesn't seem to be helpful.

Then I can't explain why the stall doesn't occur when configure.py is
launched by LyX at start. Then I am not really interested in debugging
this one as everything seems to work ok with the cygwin version of LyX...

Windows is really like a can of worms or a Pandora's box, at your
choice ;-), and I can stand it only because of cygwin.

-- 
Enrico

Reply via email to