NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/comp/os/linux/21476

Text File | 1992-12-22 | 1.7 KB | 41 lines

Newsgroups: comp.os.linux
Path: sparky!uunet!uunet.ca!xenitec!mongrel!shrdlu!gdm
From: [e-mail] (Giles D Malet)
Subject: SED is buggy (library bug ?)
Organization: 3.141592653589793238462643383279502884197169399
Date: Tue, 22 Dec 1992 19:45:57 GMT
Message-ID: <[e-mail]>
Lines: 31

I got the `sed' from the GCC disk about 3 weeks ago, and noticed it made a
mess of a couple of cnews' reports. So, a couple of days ago, I got the
SLS one, to find it is also buggy. In fact, it is the same `sed' :-(

Here is the problem :

$ echo xxxfooyyy | sed 's/x*\(foo\)y*/\1/'

should print just `foo' (ie substring 1 in the left part of the pattern) -
instead it prints the whole lot.

My old sed comes from the first release of the shared libraries,
and works fine. It currently uses libc_v2_11c. As a test I linked its
intermediate file ( .a ? can't remember) with the new libraries to form a
static module. This also produced the buggy output, which is what makes
me think something is wrong in the libraries.

This is gcc 2.2.2d, jump libc.so.4.2, and its static libs.
OS is v0.99 with tcp patches, but this failed on 0.98P5 as well.

Anyone have any ideas ? If someone can tell me where I can find the
source for sed, I will pursue this further myself.

BTW - the `newsdaily' reports in cnews are the only things I have noticed
so far that are being upset by this bug, despite a pretty busy system, so
it appears `safe' to use the buggy one. Either I do, or I get `old stat()
call' msgs all over the place...

--
Giles D Malet                                             [e-mail]
Waterloo, Ont, Canada       [phone]       [e-mail]