I came across a configure script that used this test:
echo 'alma,korte,banan' | grep -oEia ',K[^,]+,'
and expected the result to be
,korte,
That looks fine:
-o prints only the matching part from the line
-E selects egrep syntax, which allows the + operator
-i makes it case-insensitive, so that ,k matches ,K
-a makes grep assume the input is ascii, which is a no-op here.
However, on 9.3 it prints nothing.
On 10.0_RC3, the same.
By trying simpler versions, it appears it is the combination -o -i that
causes the problem. If you leave out -i, and adjust pattern and input to
have matching case, then the test case works.
Another simplification, omitting -E and replacing + by *, has no effect.
If you use another way to find what substring was matched, by replacing
-o with --color=always, you find similar results. Interestingly, this:
echo 'alma,korte,banan' | grep -Eia --color=always ',K[^,]+,'
does actually print the input line as a match, but nothing in it is
coloured as the match.
I built (on the 9.3 system) the source from -current src/usr.bin/grep,
and there it works. This confused me for a while - then I discovered
that this isn't the version in actual use. That's the one in
src/external/gpl2/grep.
The GNU grep from pkgsrc (I tried grep-3.11) works on this example.
So maybe all that is needed is to update our in-tree version. It is GPL
3+, though. The one we have seems to be version 2.5.1a(?), GPL 2+,
The CHANGES file from upstream https://git.savannah.gnu.org/git/grep.git
lists:
* Noteworthy changes in release 2.6 (2010-03-23) [stable]
...
grep -i -o would fail to report some matches; grep -i --color, while not
missing any line containing a match, would fail to color some matches.
related to commit 70e236167c3973fc428d2b5b297218fde9b68e73, committed
2010-03-17
Unfortunately this is a rather large commit, due to multibyte support. I
expect that some of the changes are not directly related to this bug,
but the parts that clearly are, are still not trivial. So I expect that
patching just this bug isn't trivial, and simply importing a recent
version from upstream is to be preferred. Ours is very, very old...
-Olaf.
--
___ Olaf 'Rhialto' Seibert
\X/ There is no AI. There is just someone else's work. --I. Rose
signature.asc
Description: PGP signature
