RE: Intermittent Issues with RCS binary co

[Meinert, Debora L.]

See answers in RED below.

-----Original Message-----
From: Paul Eggert [mailto:address@hidden
Sent: Thursday, April 03, 2003 3:12 PM
To: Meinert, Debora L.
Cc: 'address@hidden'
Subject: Re: Intermittent Issues with RCS binary co


"Meinert, Debora L." <address@hidden> writes:

>  co -r1.0 "XXGLPMCA.rdf" > "XXGLPMCA.rdf.out" 2>&1
> 
> We find that the file, that has been checked out, gets truncated by more
> than half the bytes of the original file.  We are running on an HP-UX
> version 11.11 Superdome server, utilizing MC/ServiceGuard (failover).

What kind of filesystem is storing the files?  I'm not familiar with
MC/ServiceGuard.
The filesystems are VXFS (journaled filesystems).  MC/ServiceGuard is HP's
high availability software. (server failover) 

Are the RCS files themselves OK?  That is, do they contain all the
data in question?
Yes, as far as we can tell.  The filesystem that contains all our RCS files
is .../.../archives and from what we can see the files are in tack there.
They get corrupted after the checkout command. (above)

> Have you had anyone report a problem similar to this

No.  Many people use RCS under HP-UX, though I haven't had any
specific reports for HP-UX 11.11.

> can you offer any suggestions?

If the problem is reproducible, then I'd use the "tusc" command to see
which system calls are being executed.  This should help you isolate
the problem.  <ftp://ftp.cup.hp.com/dist/networking/tools/tusc.7.5.shar>

One possibility is that it is a bug in the HP-UX 11.11 implementation
of mmap.  If that's the case, you can work around the bug by manually
modifying RCS's configuration file so that RCS thinks that mmap is not
supported.  Then remove all the .o files and type "make" again.  This
will hurt performance, but it would work around an mmap bug if that's
what's causing your problem.
Because it is intermittent, it is not something we can test at this time.
When it happens we can generally reproduce it using the same files.  We will
try this next time it occurs.  I don't think we have the luxury of removing
and recreating our files.  This is a production system with very much
visibility and performance degradation is not an option.  But we'll try to
figure out the cause first. 

Any other suggestions while we wait.  

Thanks.