bug#23361: 【Bug】bug report of GNU grep

[Eric Blake]

tag 23361 notabug
thanks

On 04/24/2016 08:04 AM, 谢敬锋 wrote:
> 
> Hi all,
> 
> Suppose the file content is as below:
> 
> abc.h
> 
> hello world
> 
> 
> 
> 
> the output of grep "*.h" file and grep -E "*.h file" are different,

Correct, and this is not a bug.

POSIX defines two different flavors of regular expressions: basic (when
you use 'grep' without -E) and extended ('grep -E'):

http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap09.html

> '*' is a regular expression meta-character.

But when it appears as the first character of a regular expression, it
had different meanings.  Read what POSIX says:

For a BRE, in 9.3.3:
"
*
    The <asterisk> shall be special except when used:

        In a bracket expression

        As the first character of an entire BRE (after an initial '^',
if any)
"

which means that as written,
grep "*.h" file

is looking for a LITERAL star character followed by the '.'
metacharacter for any character followed by a literal 'h'.  Your example
file did not contain that pattern.

For an ERE, in 9.4.3:

"
*+?{
    The <asterisk>, <plus-sign>, <question-mark>, and <left-brace> shall
be special except when used in a bracket expression (see RE Bracket
Expression). Any of the following uses produce undefined results:

        If these characters appear first in an ERE, or immediately
following a <vertical-line>, <circumflex>, or <left-parenthesis>
"

which means you have undefined results according to POSIX, and therefore
we can make it mean whatever we want, including ignoring the invalid
"*", and searching for the regular expression ".h" instead.  Which
explains why:

grep -E "*.h" file

has a match, and adding --color shows that the matching portion is the
".h" portion of the "abc.h" line.

> 
> Please help clarifying this issue!

Maybe you are confusing globs (where "*.h" matches "abc.h" because the
'.' is a literal character, and the "*" means "one or more characters")
with regular expressions (where "." means "any character", and "*" means
"zero or more repetitions of the previous regex construct, unless there
is no previous regex construct, in which case it is well-defined for BRE
but undefined for ERE").

At any rate, this is not a bug in grep, so I'm closing the bug report.
But feel free to add further comments or questions on this thread.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

signature.asc
Description: OpenPGP digital signature