🚀 go-pugleaf

In <830175140snz.4198@grempc.demon.co.uk>, I wrote:
>Tom Wheeley <tomw@tsys.demon.co.uk> wrote:
>>Surely it must be possible to make EXPIRE not mess up the indices when it
>>does run out of space ?  That is the main concern, really.
>
>I fixed this in some patches I sent to Malcolm a couple of weeks back 
>by re-correcting the active file numbering on a 'disk error'
 [...]
>The combination of these two patches should allow you to run expire 
>quickly and as many times as you need to free up the disk space.

Famous last words ... place the above on hold as well - Unlucky Alf 
here has just managed to trash his history.snw as there's no checking 
for disk full/write failure in expire.c expire_history(), is there?  
(well, there is now ... ;)

I've just copied the contents all of the remaining free clusters onto 
another partition but there doesn't seem to be much of it left :(
Hey ho, I can feel a 'build a new history.snw from article files and 
.idxs' util coming on ... (OK, so I can get away without history.snw 
and live with no cross-post detection and some duplicates but I'm just 
putting the finishing touches to a little util which builds an optimum 
KA9Q history from history.snw to improve news downloads ...)

[ As a postscript to this, I have managed to cobble back together 
  a 7,000 line history thanks to judicious use of PFE, sort and uniq
  and also bullet-proofed expire so it won't happen again - if anyone
  wants the diffs, complete source or compiled .exe, drop me a mail
  (Oh, if you want the exe and have an XT/286, be sure to mention it
  else you'll get the 386/optimised build ...) ]

Cheers, Graham.
-- 
----[ Graham E. Kinns ]----==========----< Graham@grempc.demon.co.uk >----