[Toybox] [PATCH] Fix dmesg -c and -C.

[Rob Landley]

On 01/04/2017 01:08 PM, enh wrote:
> I never use these, so I didn't notice I'd broken them until someone who
> does bringup complained.

Er...

$ git diff toys/*/dmesg.c | diffstat
 dmesg.c |   73
+++++++++++++++++++++++++++++++++-------------------------------
 1 file changed, 38 insertions(+), 35 deletions(-)

I'm in the process of redoing that entire file. But I'll take a look.

Alas, all of that was done during my trip to San Diego and I haven't had
a chance to look at it since. Seriously, the _second_ biggest downside
of the constant interruptions breaking my development time into small
chunks with large gaps between them is I do 2/3 of something and then
come back and spend that much time/effort _again_ working out what I
already did and what I haven't done yet and why I was doing it. Half the
time I just throw out my local changes start over so I'll at least know
where I _am_, but after the third time you do that to something like
"dd" you lose all mental model of what the code _currently_ looks like
and what didn't get saved, and what you didn't do because it turned out
to be bad idea and what you didn't do because you just got interrupted
multiple times before you could finish, test, and check it in. Sigh...
This is why my current work directory is "toybox/toy3", because
toybox/toybox and toybox/toy2 have piles of unifinished commits against
a now-stale version until I just started making changes in a clean
directory. And of course toy3 is:

$ git diff | diffstat | tail -n 1
 33 files changed, 467 insertions(+), 427 deletions(-)

> The "one weird trick" with SEEK_DATA is documented at the URL we already
> point to. SEEK_DATA was added in Linux 3.1 (2011) and isn't available in
> glibc 2.19 (2014), so I've added that to "portability.h" for the benefit
> of Ubuntu 14.04.

Another half-finished thing I have in my tree is xsendfile_sparse(), so
oddly enough I've already got a local patch adding that. :)

(Because initramfs can't do xattrs unless cpio can store them, and as
long as we're proposing a change to the cpio format it should do 64 bit
timestamps with nanoseconds and handle sparse files. I've restarted work
on that ~3 times now...)

Speaking of portability.h, I've pretty much given up on uClibc
(http://lists.busybox.net/pipermail/buildroot/2016-December/180102.html)
and am trying to figure out if I should remove the uClibc portability.h
bits or leave it there. I should probably check if buildroot's uClibc-ng
works without them, but I haven't had a chance to spend time on it yet...

> Also make -c and -C mutually exclusive.
> 
> Also fix some of the formatting I introduced earlier. (A clang-format file
> would help prevent these mistakes...)

I dunno clang-format. Is it similar to lindent?

  https://github.com/torvalds/linux/blob/master/scripts/Lindent
  https://www.freebsd.org/cgi/man.cgi?query=indent

(This is a thing I'm aware of, but have not personally found useful.
Fixing up whitespace is one of the things I do during cleanup passes. :)

Thanks for the patch, I'll try to take a look this evening,

Rob