> or you had the broken HDS that I fixed a few hours ago....
Could be. I did the original checkout yesterday.
It may still be a good idea to initialise "slot" to a safe value in
rec1_open_file before checking the inherited status.
David
>
> On Thu, 8 Sep 2005, David Berry wrote:
>
> > Peter,
> > I'm currently doing a complete CVS build from scratch on my new 64
> > bit laptop. All went well until it got to sst, when prohlp got a segvio.
> > Poking about using the debugger revealed that it was crashing in
> > rec_attach_file.c at the line
> >
> > hds_gl_64bit = ( rec_ga_fcv[ slot ].hds_version > REC__VERSION3 );
> >
> > because "slot" had a huge value. What seemed to be happening was that
> > an error had occurred previously, causing rec_open_file to return
> > without action, leaving "slot" in an uninitialised state. Maybe
> > rec_open_file should initialise a safe value for "slot" before checking
> > the inherited status.
> >
> > David
> >
> >
> > PS I tried making this modification myself, but funnily enough, the
> > problem then went away (i.. no error was reported and there was no
> > segvio). I then deleted my hds checkout and did a fresh checkout and
> > built it again. But the problem still did not recurr. Presumably I've got
> > some state information somewhere left over which is preventing the problem
> > from occurring.
> >
>
> --
> Tim Jenness
> JAC software
> http://www.jach.hawaii.edu/~timj
>
|