Dear Chuck
Thanks for sharing your cygwin-experiences! Once I finish my battle with
non-linux OSes I may give a little report as well.
It's hard to track down problems like these. I was just looking on the
web to see if it's a cygwin/tcl/tk problem (ie other tcl/tk programs
have problems with the latest cygwin as well) or an FSL problem (ie some
FSL-specific thing that works only with older tcl/tk) but that doesn't
really matter if a solution is not in sight. I guess it's the 1st one,
because building FSL with its own libraries (`freeware' option) does not
work on my cygwin either.
The story so far is quite short!
My cygwin experiences so far have been very good, especially because my
expectations were quite low... I installed cygwin from the net,
deselecting everything (which requires a strategy because depencies are
resolved on the fly!) and then only selecting the needed packages such
as gnuplot, tar, bash, tcl/tk, etc.
I got tix from somewhere else: http://tix.sourceforge.net/download.shtml
has a tcl/tk/tix archive that includes tcl/tk-8.3.4 and tix-8.1.3.
The resulting `cygwin disk' is really small, about 200M.
To not use the FSL freeware option (use cygwin's gnuplot, wish, tclsh
etc) I had to make some links from usr/local/bin to usr/local/fsl/bin.
It's getting a bit dodgy now :).
I then did the fsl.sh things (setting the environment) and ran build,
and the FSL menu came up! But then there were these mysterious errors.
I haven't run any of the tools from their interfaces, but they all come
up if I call them from the command line, ie Bet_gui, Susan_gui, etc. If
it's a Tcl/Tk thing, the important bit (the analyses) should all be
kosher I guess.
AM
Chuck Theobald wrote:
> Hello,
>
> I do not have a list of dependencies. In trying this approach, I kept
> running into such dependencies. Rather than track them all down and list
> them, I simply installed everything. I would suggest that you keep with
> your approach and keep a list of the dependencies you find, it would be
> a valuable addition to this mailing list's archives. Be aware, though,
> that FSL can fail silently and in non-obvious ways. One such failure
> occurs when newer Linux kernels report that something like 'head -1' is
> deprecated and that you should use 'head -n 1'. This is a warning, but
> FSL sees it as an error. Under cygwin, this is particular issue is not a
> problem, but missing dependencies may cause this type of failure.
>
> I'm sure you've seen on this list that Linux is greatly recommended over
> cygwin. I would second that. Even more, though, I would suggest that
> your choice of platform be consistent. It is problematic for a number of
> reasons to try using cygwin and Linux FSL to analyze the same set of
> data. Specifically, the paths and links FSL uses are not cross-platform.
> If you have multiple machines analyzing data, make them as similar as
> possible.
>
> For our write-up on travails with cygwin/FSL, see:
> http://lcni.uoregon.edu/mediawiki/index.php/SOFT:Cygwin
|