On Thu, Jan 27, 2011 at 9:11 PM, Peter O'Gorman <pe...@pogma.com> wrote: > On 01/27/2011 12:56 AM, Ralf Wildenhues wrote: >> >> I like this as a first measure. > > This is what I am pushing - it ends up looking like this in the wrapper > script: > > func_exec_program () > { > case " $* " in > *\ --lt-*) > for lt_wr_arg > do > case $lt_wr_arg in > --lt-*) ;; > *) set x "$@" "$lt_wr_arg"; shift;; > esac > shift > done ;; > esac > func_exec_program_core ${1+"$@"} > } > > Thank you!
Sorry for the going quiet, was out of town this week so didn't have a lot of free time. This definitely helps a lot for our case. Timings are not on the same machine as my original email, but on one slower machine it takes the worst case test case from 40 seconds back down to 11.5 seconds running with the libtool wrapper; the unwrapped .libs/lt-* binary takes 11.2 seconds so the overhead is drastically reduced. Thank you for looking into this! -Dan