Hi Dominik,

First of all 4.1.3 is old.  I'll make a new snapshot this week.  

This signal issues are simple, but annoying, and also occur on glibc Intel
Linux (eg Redhat 5.0).  It would be a very doable and useful target to
compile Coda on a RedHat 5.0 platform and fix the signal handling. 
Perhaps Elliot Lee <sopwith_at_redhat.com> is doing this already. 

There are much more serious problems with the Alpha stuff than the signal
handling.  The problem of packing the PRC2 and volume data structures
correctly is the hardest part. (the Usenix ;login: of December contains a
good paper explaining the issues.)

There is no Coda for HP/UX -- only RVM (a sub package for transaction
handling in the logs) can use PThreads (and that didn't work on Linux when
I tried it).  Coda is not thread safe yet, and the problems here are quite
elaborate, so we will stick with our co-routine threads for a little (in
particular also because porting them to Windows 95 and NT has not been a
problem, and there are no pthreads there). 

Porting to 64bit is on our todo list.  When we have Coda for i386 with
glibc that would be a natural next target.  Before starting we will have
to "publish"  a description of the structures we use and the sizes they
should have.  Currently too many things are "long", which is bad on LP64
systems (which is becoming the standard). 

Are you interested in joining that effort then?  If you are dying to get
going right away, let me know then I'll try to separate out some stuff
that you can do now.  In fact doing the threads package is a good idea. 
All contributions to user level Coda must keep a BSD style license, make
sure to check carefully on what you borrow from other code (if anything).
We will of course credit people who do work for us. 

- Peter -

On Sun, 7 Dec 1997, Dominik Kubla wrote: 

> 
> Hello Peter,
> 
> thought i should mention that i tried to compile CODA 4.1.3 on a Linux/Alpha
> system (egcs, glibc 2.0.5c, linux 2.1.61) and that i did not get very far.
> 
> If 4.1.3 is indead the latest revision, i will go ahead a provide you with
> a patch - if possible - or at least a detailed bug report.
> 
> Right now the BUILD fails at lib-src/mlwp with all sort of errors regarding
> incompatible types in sigsetmask et al.  The docs hint that CODA for HP-UX
> uses pthreads,  wouldn't that be better for Linux as well?  This way we could
> use the native threads of glibc2 (LinuxThreads)...
> 
> Dominik Kubla
> The FreeLinux Project
>