On 05/16/2011 12:57 PM, Jim Meyering wrote: > Paul Eggert wrote: >> On 05/16/11 07:44, Pádraig Brady wrote: >>> I tested your ls ordering trick on linux, solaris 10 and freebsd and it >>> works. >>> Please push. >> >> Thanks, done. > > Thanks. I appreciate the added comments, too ;-) > > What do you think of this addition? > It may be more efficient to invoke one subshell rather than two, > even though the new subshell includes an additional pipe to sed. > For the record, I don't really care either way, since even for coreutils, > saving 1000 subshells (one per file from gnulib) is probably a > negligible gain in performance, overall.
Anything that shaves 1000 processes is worthwhile for cygwin. However, > @@ -677,9 +677,8 @@ symlink_to_dir() > # so that broken tools aren't confused into skipping needed builds. > See > # <http://lists.gnu.org/archive/html/bug-gnulib/2011-05/msg00326.html>. > test -h "$dst" && > - src_ls=`ls -diL "$src" 2>/dev/null` && set $src_ls && src_i=$1 && > - dst_ls=`ls -diL "$dst" 2>/dev/null` && set $dst_ls && dst_i=$1 && > - test "$src_i" = "$dst_i" && > + inums=`ls -1diL "$src" "$dst" 2>/dev/null|sed 's/ .*//'` && set $inums > && Two subshells with no pipe vs. one subshell with a pipe produces the same amount of fork/exec calls under bash, so you aren't saving processes in this case. On it's own merits, I have no strong opinions about keeping or dropping this patch. But you gave me an idea that _does_ speed up cygwin: inums=`ls -1diL "$src" "$dst" 2>/dev/null` inum1=${inums%%[!0-9]*} inums=${inums#* } inum2=${inums%%[!0-9]*} test "$inum1" = "$inum2" which computes the same with only one child process, by avoiding both a second ls and a pipeline. On the other hand, it only works with POSIX shells (Solaris /bin/sh doesn't understand the XSI notion of %% and ##). -- Eric Blake ebl...@redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
