After assembling our genomes - usually with Mira, we use tg_index to make gap5 databases from Mira's CAF files.
There is a version incompatibility with gap5 databases made with tg_index on Linux computers when edited using gap5 on a Mac, except when tg_index from the pre-compiled package (staden-2.0.0b9.x86_64.tar.gz) is used. It occurs as a stack overflow (Segmentation fault 11) on the Mac.
Today, I updated the Staden package on my Linux boxes and Mac minis using today's code from svn. I keep old versions of the Staden package around in case of issues.
On the Linux boxes, gap5 compiled from svn version 3107 will edit databases created from its own tg_index (version 3107) and from the pre-compiled package (version 1.2.13-rexported). So there's likely nothing wrong with the Unix code!
If I move the gap5 database made with tg_index 3017 to any of my Macs, it crashes with a segmentation fault 11 when gap5 on the Mac is used. If I move the gap5 database made with the pre-compiled tg_index (1.2.13-reexported), gap5 on the Mac works fine.
I can repeat the same experience using tg_index from svn version 3103 on our Unix server, i.e. the gap5 database segfaults when using gap5 on the Mac.
I can repeat this with the same versions of tg_index on two different Linux boxes, and seven Macs.
So, the logical solution would be to build the gap5 database on the Mac, if the Linux version won't work. Read on.
If I use tg_index from svn versions 3102, 3013, and 3017 built on the Mac, it will not make a gap5 database, i.e. it crashes with a segmentation fault 11 when gap5 is invoked.
The load on our network is too high for everybody to use gap5 on the Linux system, which is why we use the Mac minis - they are cheap, unix, and stand-alone.
As usual, it appears to be a MacPorts issue. Has anybody tried compiling and running the Staden package using 'Homebrew' or 'fink'?
Do you know when this incompatibility (and tg_index crashing on Macs) first occurred? Again our lack of Mac development machines is a problem.