bug#20679: A bug of pwd

2015-05-31 Thread Bernhard Voelker
On 05/31/2015 04:35 AM, Bob Proulx wrote:
 Bernhard Voelker wrote:
 First of all, I want to mention that the invoked 'pwd' is a builtin
 in most shells, which means you have to e.g. specify the path like
 /bin/pwd to be sure to invoke the coreutils version of it.
 
 A very, very small comment.  This is all true but the wording makes it
 sound somewhat like a recommendation to use /bin/pwd in order to get
 the coreutils program.  I don't think that was intended.  I think
 instead it was intended only that normally the user has called the
 builtin pwd and the builtin is not coreutils and therefore nothing we
 would do about it here.

Indeed, thanks for making that clearer.

Have a nice day,
Berny





bug#20679: A bug of pwd

2015-05-30 Thread Bob Proulx
Bernhard Voelker wrote:
 First of all, I want to mention that the invoked 'pwd' is a builtin
 in most shells, which means you have to e.g. specify the path like
 /bin/pwd to be sure to invoke the coreutils version of it.

A very, very small comment.  This is all true but the wording makes it
sound somewhat like a recommendation to use /bin/pwd in order to get
the coreutils program.  I don't think that was intended.  I think
instead it was intended only that normally the user has called the
builtin pwd and the builtin is not coreutils and therefore nothing we
would do about it here.  Reports about the builtin would go to the
shell.

One can compare the builtin by using both a plain pwd version and
comparing against the /bin/pwd coreutils version in order to test
differences between them.  If they work the same then it is very
unlikely it would be a bug since they are independent implementations.

For typical scripting it would be normal to continue to use the plain
pwd and use the shell builtin version.  But proper shell quoting is
still needed.  :-)

Bob





bug#20679: A bug of pwd

2015-05-28 Thread Bernhard Voelker

tag 20679 notabug
close 20679
stop

On 05/28/2015 06:08 AM, 徐兴旺 wrote:

If the name of current diretory including blank, pwd return raw blank istead of escaping 
character, \ .
When used with cp command, the directory name will be recognized two params.


If I understand correctly, then you have a script doing something like
the following:

  $ mkdir a dir
  $ cd a dir
  $ cp -v /dev/null $(pwd)
  cp: target ‘dir’ is not a directory

First of all, I want to mention that the invoked 'pwd' is a builtin
in most shells, which means you have to e.g. specify the path like
/bin/pwd to be sure to invoke the coreutils version of it.

Second, the output of pwd(1) is correct - i.e., there is no such
thing than escaping characters.  Instead, the caller has to care
about proper quoting:

  $ cp -v /dev/null $(pwd)
  ‘/dev/null’ - ‘/tmp/a dir/null’

Therefore, I'm marking this as not a bug in our software.  But in case
of questions please don't hesitate to discuss this further by replying
to this bug.

Have a nice day,
Berny







bug#20679: A bug of pwd

2015-05-27 Thread 徐兴旺
If the name of current diretory including blank, pwd return raw blank
istead of escaping character, \ .
When used with cp command, the directory name will be recognized two params.