Re: bash hash bug

[Bob Proulx]

Linda Walsh wrote:
> Erp...  I probably thought it was getting rid of the problem I've had
> in the past with ksh-type shells of retaining a bad-mapping to a file that
> was no longer accessible.  A minor annoyance, admittedly, but one that
> entangled me a few times several (or more) years ago...

You probably were.  Because that is exactly why I have it set in my
environment too.  Otherwise if files are moved around bash fails to
find them in the new location.  So checkhash is just the behavior that
I want.  Of course I could just learn to use the csh-like 'hash -r'.  :-)

> Chet Ramey wrote:
> > When this option is on, bash has to check the hashed filename for
> > each hash lookup.  This essentially causes the hash entry to be
> > deleted and re-added each time, which resets the number of hits to 1.
> > (It could probably be done without the deletion and re-addition; I
> > should look at that.)

So there is the root of the current behavior.  I know that I had never
noticed that before because I never dumped out the hash tabled.  But
just because there is a deletion and addition I don't think this is a
performance issue.  (Is it?)  And so the count is purely cosmetic.

Did you have a purpose for the count other than just double checking
that the hash was working?

> > (This is not to say it could not behave better -- it can.  But
> > it's disabled by default.)

I read in the above that there is agreement with you that there could
be improvements there, but that the current behavior is it is at the
moment.

Bob