Re: shell-command - missing shell-quote-argument for program?

[Lennart Borgman <lennart.borgman.073@student.lu.se>]

Dave Love wrote:
> Lennart Borgman <lennart.borgman.073@student.lu.se> writes:
>
>   
>> Thanks for responding. I am sorry if I were not clear enough about the
>> problem. The problem is that when you call shell-command (or
>> shell-command-to-string) the arguments may need to be quoted so that
>> the shell does not mistreat them.
>>     
>
> ??  Shell-quoting the args would break them unless they comprise a
> single shell word, in which case there's no point.
>   
There must be some misunderstanding here, but please let us not discuss 
this now. I think we have resolved most of the issues I have been 
complaining about here.

>   
>> A common example is file names with
>> spaces in them. If a file name with spaces are used as argument to
>> shell-command it must be quoted. Otherwise the shell will treat it as
>> several arguments.
>>     
>
> Well, that's different; maybe that's what you meant by the first bit.
> If you're composing a shell command you have to DTRT, but
> `shell-command-to-string' is probably the least of your problems with
> file names that shells split.  You shouldn't normally use
> `shell-command' non-interactively.
>
> Why don't you submit patches?  It's still not clear to me what you're
> complaining about.  There are surely more serious Emacs issues to work
> on, though.
>   
These are possible bugs and I consider them serious. The most common 
problem is that with file names containing spaces.

>   
>> I found examples of what I suspected was missing quoting in the files
>> I mentioned and I therefore mailed you as authors (since there is no
>> maintainer mentioned in the files).
>>     
>
> If you sent to me because of python.el, the maintainer is supposed to
> be FSF, unfortunately.  (I assume you mean the version in the Emacs
> sources rather than the fixed and enhanced one I maintain.)
>   
Yes, but I wanted your opinion only.

>   
>> - emerge.el
>> According to Dave it is probably obsolete and replaced by ediff. Then
>> I will just leave emerge.el as it is.
>>     
>
> If it's not in lisp/obsolete it still needs to be maintained.  It's
> also documented in the manual, rather than Ediff, unfortunately.
>   
Looks like you are right.

>   
>> - progmodes/python.el: python-after-info-look: python-command
>>     
>
> I don't have the Emacs CVS version to hand, but if that has a problem,
> please chase whoever changed it.  In my version the only instance of
> `shell-command' is:
>    (shell-command-to-string (concat python-command " -V"))
> and I don't understand what you mean by referring to python-command.
> Have you read its doc (or has that been changed too)?
>   
My impression right now is that this special case could not be quoted 
since python-command could be a file name with arguments.

> The thing that is/was wrong in principle with command arg processing
> in python.el is `python-args-to-list' -- see its fixme.  Similarly for
> anything else that's copied the mechanism from cmuscheme.el.  (The
> inferior interpreter mechanism is one of the things which should be
> abstracted, though, rather than fixing individual versions IMNSHO.)
>   
Thanks for the pointer. Has this been fixed in the version of python.el 
that you maintain yourself?