Re: 1.11.[2&9], server abort, "unable to parse"

[Todd Denniston]

Larry Jones wrote:
> 
> Todd Denniston writes [using very long lines]:
[sorry, did not want to clip cvs error messages. back to 72 now :]
> >
> > Short: I think something bad happened between 1.11.1p1 and 1.11.2 to
> > the rcs file parsing on loaded machines.
> 
> The critical change that happened between 1.11.1p1 and 1.11.2 was that
> we started using mmap() to access the RCS files when possible.  It would
> appear that mmap() exists on your system but is seriously broken.  You
> should check to see if there are any OS patches to fix it.  You might
> also want to try building CVS with the Sun compiler rather than GCC.  As
> a workaround, you can undefine HAVE_MMAP in the generated config.h to
> force CVS to use regular file I/O instead of mmap().
> 
> -Larry Jones

yep, if I make the following change, and point the server side at its
binary, everything seems to work correctly with my test cases.

diff config.h.normal config.h
199c199
< #define HAVE_MMAP 1
---
> /*#define HAVE_MMAP 1*/

is there an option I can pass to ./configure that will set it this way,
so I only have to keep the config.log around to know what I did to the
last version of cvs I compiled?



I kind of wonder how during the year I used it on a small project
(<30MByte) I never noticed a problem with 1.11.2....


 From doing a bit of searching it looks like Sun does have some patches
that mention mmap problems in the Solaris 2.6 series, however I a don't
recognize any of the descriptions as necessarily the problem.

http://sunsolve.sun.com/pub-cgi/retrieve.pl?doc=fpatches%2F106625&zone_32=mmap
http://sunsolve.sun.com/pub-cgi/retrieve.pl?doc=fpatches%2F105693&zone_32=mmap
http://sunsolve.sun.com/pub-cgi/retrieve.pl?doc=fpatches%2F105181&zone_32=mmap


there is only one of these that I am not finding some version of
installed on the machine.

-- 
Todd Denniston
Crane Division, Naval Surface Warfare Center (NSWC Crane) 
Harnessing the Power of Technology for the Warfighter