[BiO BB] Need installation of Bio software on Mac

[Corentin Cras-Méneur]

À 10:14 +0200 le 06/15/05, Dr. Christoph Gille a écrit:
>Dear Corentin,

	Hi Christoph,

>many thanks for taking the time to try this out.


Not a problem.

>I generally append the suffix  dot exe to binaries because this allows me
>to have the same documentation for DOS and UNIX.

I see.


>A symbolic link (ln -s ...) of the clustalW or dialign-t binarie to the
>specified directory is only recognized when it gets the ending dot exe.
>Most binaries like clustalW come in an  JAR archive because some require
>more than one file. Even though the clustalW version is the ordinary
>C-version and not a java-version.


Very true too (though you can have a system to 
detect the platform and adapt the symlinks 
accordingly, but I get your point).

I still believe it could be worth querying the 
system to check whether the apps are not already 
installed somewhere else (eg: /usr/local/bin), 
but I realize it represents more work for you.

>
>I am a bit disappointed that the installation of dialign-t fails on Mac.
>
>Theroetically, STRAP should go to the directory
>$HOME/charite.christo.settings/bin/dialign-t/
>
>Subsequently the command make
>should be executed
>
>finally the binary should be renamed
>mv dialign-t dialign-t.exe
>
>
>Do you know which of these steps failed ?


The console reports:
>2005-06-15 18:19:02.536 java[692] tabletProximity:
>FocusManager noResponderFor called
>2005-06-15 18:19:18.955 java[692] tabletProximity:
>Jun 15 18:19:31 Corentin java[695]: tabletProximity:
>Jun 15 18:21:26 Corentin java[695]: tabletProximity:

The .jar file was properly downloaded and the 
dialign-t folder contains all the source files.
I (very briefly) saw a warning that mentioned 
make (but it was too fast nor me to actually read 
it).

If I move to this directory and run make I get:
>gcc  -O3 -march=i686 -funroll-loops  -c -o museq.o museq.c
>cc1: error: invalid option 'arch=i686'
>make: *** [museq.o] Error 1


If I rename Makefile.MAC_OS into Makefile and run 
make, I get the binary. I can then rename it and 
use it in the app.
I used it on the example sequences and all I got 
was "no score availabel" (with the same typo). 
Since I know almost nothing about dialign-t, that 
might as well have been expected.



>Supposed it would work, how do you like the concept of distributing the
>source rather than executables ?

It's a tough choice.
Binaries are convenient and allow deployment 
regardless of the fact that people have installed 
gcc.
Distributing the sources instead would require 
that people have installed X-Code on their Mac 
(which should be the case for most people running 
command-line applications for bioinformatics 
anyway).
I tend to like the sources better...


Don't hesitate to ping me off-list if you need 
additional information and don;t want to bother 
the rest of the list.


	Corentin