grephistory confusion

bill davidsen davidsen at tmr.com
Tue Nov 5 20:14:23 UTC 2002


In article <Pine.LNX.4.44.0211042254490.11083-100000 at puck.litech.org>,
Jeffrey M. Vinocur <jeff at litech.org> wrote:
| 
| On Mon, 4 Nov 2002, Russ Allbery wrote:
| 
| > Jeffrey M Vinocur <jeff at litech.org> writes:
| > 
| > > I just rebuild history (but not overview) from the spool.  I picked an
| > > article at random and saved its Message-ID and storage token from before
| > > the rebuild.  Now using grephistory on that Message-ID claims that it's
| > > not in history, but if I grep for the token it definitely does exist
| > > (and I can retrieve the article correctly using sm on the token).
| > 
| > Sounds like the history indexes (generated with makedbz) are corrupted to
| > me.  Either that, or the message ID hash wasn't done right.  A grephistory
| > option to just print out the hash of the message ID would be useful here,
| > so that you could grep the history file for that hash and make sure that
| > part is working properly.
| 
| Ok, that sounded interesting.  Just committed to CURRENT.
| 
| And...the hashing seems to be being done correctly, but grepping history
| manually for that hash is unsuccessful.  So makedbz is working fine, and
| the -O part of makehistory is working fine, but somehow the bit that
| generates the history database isn't??
| 
| This is getting weirder -- I just started taking tokens out of history and
| feeding them to sm, and I'm getting something really crazy.  I have here a
| token, which I got out of history, but the Message-ID found in the article
| I get by calling sm on that token does *not* hash to what's on the line in
| history that had the token!
| 
| (If that was so convoluted you can't follow it, let me know.)

Jeff, if you're hacking about in grephistory, maybe you could add a
feature for me as well. I occasionally want to generate an innxmit input
file from a list of message-ids. Now if I do them one at a time on the
command line I have the MID and the token and can write both to the
file, saving the overhead of fetching the token to get the MOD.

If I just run the message-ids in stdin I can use -s and do it all in one
process, but the message-id gets omitted and there are often missing
messages (cancels).

If there were an option to emit both, say "-S", that would be useful. I
have been planning this for over a year and never seem to find the need
justifies the time. Since you are up to speed on the code and did one
thing for fun I thought I'd at least mention it ;-)
-- 
bill davidsen <davidsen at tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.


More information about the inn-workers mailing list