[Toybox] [Issues] - grep

[Rob Landley]

On Thu, Aug 28, 2014 at 6:22 AM, Ashwini Sharma
<ak.ashwini1981 at gmail.com> wrote:
> Hi List,
>
> While verifying the command grep and analysing/comparing the behavior
> with GNU and busybox. I notices following

My local copy still has the -ABC addition pending, by the way. (As in
my local copy of the code does not currently compile and is full of
debug print statements from where I left off trying to make that work.
:)

> 1. __-H__ and __-h__ options, we prefer -H over -h, GNU and Busybox prefer
> the last one mentioned on command line.

Easy enough to fix ("[-Hh]" at the end of the optstring).

> 2. links are not followed for doing grep, i.e. grep "search" <linkToFile or
> Dir> -r
>     no results are shown. Shouldn't the link be followed.

Same issue as cp's -HLP (commit 1388). Easy enough to fix, the trick
is determining what the right behavior should be. (Alas, posix's grep
standard is so utterly out of touch it doesn't even mention -r let
alone symlink behavior when recursing.)

I'm guessing the equivalent of cp -H would be correct? (Follow command
line symlinks, but not ones found while recursing.)

> 3. searching in binary files, toybox gives very minimal or no results, e.g.
>     grep "toy" toybox_unstripped, reports 4-5 lines whereas GNU/BBox fills
> the scroll       buffer.

This is probably the embedded NUL byte issue. (Standard C string
manipulation says a string ends at the first NUL byte, but here we're
getting reading "to the next newline or end of file", and that gives
us strings with embedded NUL bytes, and then we ignore everything
after the first NUL.)

Fixing this was actually on my todo list for implementing sed, then I
need to go back and add the corrected infrastructure to grep. (I also
need to fix musl's regex engine to allow "\|" to work in non-extended
regular expressions. We use that in grep to glue together multiple
search strings, but when building against musl doing that doesn't
currently work. I tracked it about halfway down last month... a couple
days before my house got flooded. Kinda dropped the ball in all the
distraction, that's also when my most recent stab at the grep -ABC
stuff got stalled...)

> The spec http://pubs.opengroup.org/onlinepubs/9699919799/utilities/grep.html
> doesn't mention anything for these behavior  or options.

Sigh. Tell me about it. :(

(I'm on the Austin Group's mailing list, that's the people who update
the posix spec. I keep feeling I should bother them about these
things, but it's down on the todo list below getting the test suite
fluffed the rest of the way out. Speaking of which, I'm still
debugging Divya's contributions...)

> Do have a look and give your inputs on
> What is the right one to be followed?

The first two are easy and I should be able to fix them tonight. The
third requires new infrastructure to be written, probably not in scope
for the upcoming release but I can put it on the todo list for the
next one.

> Thanks,
> Ashwini

Thanks for the bug report,

Rob

 1409251629.0