🚀 go-pugleaf

> Date: Wed, 3 Nov 93 14:00:13 PST
> From: dbaillie@darwin.mbb.sfu.ca (David L. Baillie)
> To: matthews@greengenes.cit.cornell.edu
> Subject: Re:  ACEDB vs. Linux
> 
> No, what version of Linux are you running?  And what release of ACeDB.
> we have not put 1-10 on linux yet, as we cannot get the dataset to 
> assemble, we are not sure where the problem lies, I gather Solaris 2.0
> chokes on the new c. elegans data release also, so Linux is not alone.
> I will forward your message to Ken Clark, who was the person who
> did the port, he may be able to suggest something.
> 
> dave

Well, I fixed my problem.  Don't know if it's related to yours or to the 
Solaris problem, but it looks like the kind of thing that might be.

Fix:  In models.wrm, change "Text" to "?Text" globally.

Diagnosis:  Apparently Linux can't handle long lines (e.g. 200 characters) as
Text, only as ?Text.

Symptoms:  xace exits with a segmentation fault when an object containing 
a long Text line is picked, or when an attempt is made to delete the object,
or to load another object that links to it.

Notes:  
1. The problem does not occur on the Sparc (xace 1.91 and 1-10), or SGI or
   DEC (xace 1.91; 1-10 not tested).
2. Freshly loaded objects usually can be manipulated without a problem.  The
   program must be exited and restarted before the failures start happening.
3. The smallest dataset that was demonstrated to fail was two objects, not
   linked to each other.

Versions and configurations:  
 xace 1-9 or 1-10, CACHE1 800 to 10000, CACHE2 256 to 2000, DISK 10000 to 20000
 Linux 0.99 patchlevel 13
 8 MB RAM, 16 to 42 MB swap
 machine Texas Instruments 486DX2/50 MHz


Dave, Ken, Jeff, thanks for your help in working on this.

- Dave Matthews