(old)Puppy Linux Discussion Forum

jamesbond wrote:

technosaurus wrote:

jamesbond wrote:Attached is a patch that will make bb_guess_fstype

I bet Denys would accept it as a replacement for blkid if we added UUID and label - they are just at different offsets and most seem to already be documented in the source already.

I'm not so sure, I thought that's what blkid is for. The code in volume_id is actually meant for blkid, and in all the filesystem probes they actually stores both the label and the UUID if available (in fact, some of the unused probes seem to work right but they don't have UUID and labels identified properly and my guess is that why Denys left them out atm). I would be happy to drop guess_fstype altogether if only busybox blkid can specify which device to probe (instead of probing all) - but alas they are not.

Ibidem wrote:Do you mean like "busybox blkid /dev/sda1" ?
If so, it works with busybox 1.21.
But if you mean like "busybox blkid /dev/sda", that doesn't work; you must use "for d in /dev/sda*; do blkid $d; done"
blkid does need to print at least TYPE and LABEL as well to match the standard version.

jamesbond wrote:

Ibidem wrote:Do you mean like "busybox blkid /dev/sda1" ?
If so, it works with busybox 1.21.
But if you mean like "busybox blkid /dev/sda", that doesn't work; you must use "for d in /dev/sda*; do blkid $d; done"
blkid does need to print at least TYPE and LABEL as well to match the standard version.

Actually, I was wrong and you are right - even my current 1.19.4 busybox already does that

The real problem is this: if I pass a devpath for something that blkid doesn't understand (try passing an "extended partition" device)
a) util-linux blkid will produce no output.
b) busybox blkid will dump output for *ALL* devices.
Of course we can get around that in scripts, but that's *annoying*.

musher0 wrote:To whom it may concern:

At least reintroduce the prompt function and the command function in busybox's less (-P and !<command> parameters).

Or remove less from busybox altogether. Because, as it is now, busybox's less is a joke.

It's a shame that such an excellent pager/reader with so many useful options be reduced to not even a workable skeleton of itself in busybox.

It's also misrepresentation, on an ethical level, but that is another subject.

musher0

Ibidem wrote:With busybox 1.21.0, it will produce no output (tested with /dev/sda4)
This is proper behavior, since that device has no UUID.
Also, it seems LABEL is now supported. TYPE, however, isn't.

Hmm...I wonder if an mdev hook could use blkid for /dev/disks/by-*

And I believe busybox has code to detect filesystems, but good luck finding it.

jamesbond wrote:

Ibidem wrote:With busybox 1.21.0, it will produce no output (tested with /dev/sda4)
This is proper behavior, since that device has no UUID.
Also, it seems LABEL is now supported. TYPE, however, isn't.

Well It wasn't obvious from the code. Perhaps it's time to upgrade I'm quite sure it supports type because in 1.19.4 it does show type. Perhaps you didn't enable a compile-time option.

Hmm...I wonder if an mdev hook could use blkid for /dev/disks/by-*

Probably.

And I believe busybox has code to detect filesystems, but good luck finding it.

It does, in util-linux/volume_id. That's what my patch is for to make a standalone executable out of these filesystem detection code.

commit e8cfc3f693e8a0cbfc489564c76e00c074f63066
Author: Denys Vlasenko
Date:   Sat Mar 3 15:09:07 2012 +0100

    blkid: if parameters are given, do not scan /dev
    
    Signed-off-by: Denys Vlasenko

#!/bin/busybox ash
#Assume $MDEV is the device name
#Make /dev/disk/by-* symlinks

if [ -z "$MDEV" ]
then 
 echo 'Use from mdev/conf'; env >>/var/log/mdev-blk.log
# This expands to "export" when no output, or to a sequence of assignments
else
 export `blkid "/dev/$MDEV" |cut -d : -f 2` >/dev/null|| true
 [ -n "$LABEL" ] && mkdir -p /dev/disk/by-label && \
 ln -sf "/dev/$MDEV" "/dev/disk/by-label/`echo $LABEL|tr -d '"'`" || true
 [ -n "$UUID" ] && mkdir -p /dev/disk/by-uuid && \
 ln -sf "/dev/$MDEV" "/dev/disk/by-uuid/`echo $UUID|tr -d '"'`" || true

 exec >>/var/log/mdev-blk.log
 echo "blkln $@"
 env
 echo ===
fi

jamesbond wrote: Anyway, as it is now, busybox guess_fstype takes 2.5 times the time of the original guess_fstype , but I'll take use it anyway because it is still in milliseconds range and it is more maintainable. I am re-posting the patch because the original one I posted was buggy (forgot to return pointer value in xzalloc - but it worked previously as long as I didn't turn on optimisation ...).

Ibidem wrote:

musher0 wrote:To whom it may concern:

At least reintroduce the prompt function and the command function in busybox's less (-P and !<command> parameters).

Or remove less from busybox altogether. Because, as it is now, busybox's less is a joke.

It's a shame that such an excellent pager/reader with so many useful options be reduced to not even a workable skeleton of itself in busybox.

It's also misrepresentation, on an ethical level, but that is another subject.

musher0

From "reintroduce" I get the impression that you are saying it was in busybox at one point...if that is correct, when was that? (I couldn't find any indications in the git log.)

By the way, for my own purposes, (current) busybox less is quite adequate in a way that more falls far short of.

Ibidem wrote:Just for reference: here's a "guess_fs" patch based on the previous patches that provides an applet.
I did a bit of cleanup and used the //config: //applet: //kbuild: approach, which reduces the risk of patches getting stale.
Also, I added support for multiple device names.

musher0 wrote:

Ibidem wrote:

musher0 wrote:To whom it may concern:

At least reintroduce the prompt function and the command function in busybox's less (-P and !<command> parameters).

Or remove less from busybox altogether. Because, as it is now, busybox's less is a joke.

It's a shame that such an excellent pager/reader with so many useful options be reduced to not even a workable skeleton of itself in busybox.

It's also misrepresentation, on an ethical level, but that is another subject.

musher0

From "reintroduce" I get the impression that you are saying it was in busybox at one point...if that is correct, when was that? (I couldn't find any indications in the git log.)

By the way, for my own purposes, (current) busybox less is quite adequate in a way that more falls far short of.

Hello, Ibidem.

"Re-introduce" in the sense of putting back in the busybox less some of the functions it removed from the real less. I don't think those functions mentioned above were ever in the busybox less.

The difference can be illustrated by the size of the --help :
if I remember correctly : busybox less -> 4 lines
the real less -> 4 pages.

In my mind, it's clear that the busybox less is mis-representing the real less. Is there a doc telling the user what functions busybox has removed from the utilities it is supposed to replace? If not...

BTW, if you look in the /bin directory, you'll see quite a few busybox functions that BK himself has marked as "BB-NOT-USED". And he then provided the real program or a suitable replacement.

The problem here is that the user is not made aware of what (s)he is missing. If I had not searched the Internet, I would still not know all the possibilities of the real less. But not all users are as persistent.

You never know when you will need the full version of the utility replaced by a busybox "applet".

Best regards.

musher0

scsijon wrote:I sugest that you have a look at toybox and see what they have needed to added in there to extend busybox's functions.

$ ./toybox 
basename bzcat cal cat catv chgrp chmod chown chroot chvt cksum clear 
cmp comm count cp cut date df dirname dmesg dos2unix du echo env expand 
expr false find free groups head help hostname id ifconfig insmod 
kill killall link ln login logname losetup ls lsmod lsusb md5sum mdev 
mkdir mkfifo mknod mkswap mktemp modinfo mountpoint nc netcat nice 
nohup od oneit passwd patch pidof printenv pwd readahead readlink 
realpath rev rm rmdir rmmod seq setsid sha1sum sleep sort stat swapoff 
swapon switch_root sync tac tail taskset tee time touch true truncate 
tty uname uniq unix2dos unlink unshare uptime usleep uudecode uuencode 
vconfig vmstat w wc which who whoami xargs xzcat yes 

scsijon wrote:thank you ibidem,

I was was leading to the idea that maybe we should add toybox to our 'standard puppy' like aborigonal and others have done to extend some commands without having to use the 'full' command due to lack of functionality with the current busybox's version of a command. It may even allow us to remove some of the full commands we use now.

Alternately, the busybox group maybe need to revisit it's command set and extend a number of them with what users actually now need to have.

My apologies technosaurus for taking your ropic off-topic.

regards to all

Karl Godt wrote:Am a fan of busybox modprobe , its missing the -D --show-depends option i think . I hate modprobe[-FULL] !

Some are scripts to adjust vars ie reboot, poweroff, df , calling busybox applets in them .