httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ralf S. Engelschall" <>
Subject Re: [PATCH] accelerator
Date Fri, 24 Dec 1999 14:19:22 GMT

In article <199912220157.SAA24108@sanitas> you wrote:

> On OS/390 the overhead for fork and exec is extremely high.
> this make "configure" painfully slow.  Most of the time is
> consumed by "", which executes 2 or 3 "sed"s and
> an "ln -s" for each symbolic link created.  This patch replaces
> (except for a few comments).  Much of the filtering
> by "sed" is replaced by predicates in the "find".  An "awk"
> script then batches the commands so a single "ln -s" is issued
> for each directory.  This roughly halves the time taken by
> "configure" on OS/390, and slightly improves performance on
> Solaris.

Hmmm.. IMHO a slight performance increase is not worth perhaps breaking any
portability issues. mkshadow was written with 100% portability in mind and
already proved to be portable in the past and so I'm not very happy about a
completely new script. Especially because...

1. I'm not convinced that `` VAR=`pwd` || { cmd; cmd } '' is portable enough.
2. I know that `` mkdir -p '' doesn't exists on all platforms.
3. I'm not convinced that `` find ... -prune -o ! -name x ! -name x ... ''
   is portable enough.
4. I know that ``|\n## cat...\n"$SHELL"'' is hated by some shells
   (they don't continue over comments).

And I know that the current mkshadow script already costed me a lot of
iteration to make it actually work, so I'm not convinced that your script
actually does not introduce new subtle problems. At least in my review I
wasn't able to decide whether it works correctly for all cases or not. Hmmm...
so I personally vote -0 for this change until someone other proves me that we
don't introduce new problems by such a performance change.

                                       Ralf S. Engelschall

View raw message