Re: Status of install-info

[Ian Zimmerman]

Ian> So, if I just process every (appropriately named) file in the
Ian> directory, I'll end up with duplicate entries.

Eli> My failing memory suggests that install-info removes duplicate
Eli> entries.  Am I right (I don't have time to look in the code right
Eli> now)?

Well, if you're adding just one entry, it doesn't really hurt to use
linear search.  I'm regenerating from scratch, so that would be
quadratic behaviour for me.

Ian> This can be avoided with a search data structure like a balanced
Ian> tree or hash table, but that seems _really_ overkill.  OTOH I can't
Ian> just ignore symlinks, they could point somewhere else
Ian> (e.g. /usr/share/<package>) and then I'd miss their entries.
Ian> Suggestions?

Eli> How about if you ignore only those symlinks whose targets are in
Eli> the same directory?

Have you read the Linux realpath(3) manpage lately? :-) Also,
subdirectories would have to be treated the same way, and detecting that
a directory is a sub of another is a headache too with the constraints I
have (pre-standard C, needs to work on Cygwin).

-- 
A true pessimist won't be discouraged by a little success.