On Fri, Jun 01, 2007 at 08:36:19PM +0200, Souf wrote: >Le vendredi 01 juin 2007 à 19:16 +0200, Bernhard Fischer a écrit : >> On Fri, Jun 01, 2007 at 07:08:16PM +0200, Souf wrote: >> >Le vendredi 01 juin 2007 à 15:57 +0200, Bernhard Fischer a écrit : >> >> >> recursive_action() anyone? >> > >> >I had already sent a large patch to improve mdev but it does not seem to >> >interest many people, in any case I resend it here, it uses: >> > >> >xmalloc_getline() >> >last_char_is() >> >recursive_action() >> >bb_make_directory() >> >remove_file() >> >bb_get_last_path_component() >> >xmalloc_readlink_or_warn() >> > >> >all that for only 5.9 KB >> >> 5.9 Kb as in adding 5.9 KB ? > >total size of mdev after compilation = 5.9 KB >> >> What are the size(1)es for this large patch? >> > > >"large" patch = 10.7 KB >but it modifies the operation of mdev deeply:
You didn't paste the output of size(1) AFAICS. Doubling the size of mdev is not exactly a thing that strikes me as a good idea, conceptually. > >(mdev.conf) >1- create symlinks > <symlink> -> <target> > stderr -> /proc/self/fd/2 > XOR -> null Doesn't sound like it would warrant more than let's say 200 bytes. > >2- create (sub-)directories > <[path/]directory> <octal permissions> > pts 755 > shm 777 > bus/usb/001 0755 Doesn't sound like it would warrant more than let's say another 200 bytes. > >3- create a device in a (sub-)directory > <[path/]device regex> <uid:gid> <octal permissions> > snd/seq 0:0 0666 > snd/pcm[CD0-9cp]* 0:0 0666 > input/event[0-9] 0:0 0600 Doesn't sound like it would warrant more than let's say another 100 bytes. > >4- execute an command (exactly like the 1st mdev) If it's exactly like the current mdev, no size-increase warranted. > <[path/]device regex> <uid:gid> <octal permissions> [<@|$|*> <command>] > > snd/controlC[0-9] 0:0 0666 @ echo "hello $MDEV" > snd/seq 0:0 0666 $echo "goodbye $MDEV" > snd/timer 0:0 0666 * echo "Hi $MDEV" > > @ Run after creating the device. > $ Run before removing the device. > * Run both after creating and before removing the device. > >5- work with AND without CONFIG_SYSFS_DEPRECATED (kernel >= 2.6.20) That could cost a little bit, i admit, but not too much either. What's the bloat-o-meter output against pristine trunk and the size(1) output of it? _______________________________________________ busybox mailing list [email protected] http://busybox.net/cgi-bin/mailman/listinfo/busybox
