execv and then socket: "operation not permitted"

Corinna Vinschen corinna-cygwin@cygwin.com
Tue Jan 11 13:05:00 GMT 2005


On Jan 10 22:54, Eric Hoffman wrote:
> On Sun, 9 Jan 2005 13:19:53 +0100, Corinna Vinschen wrote:
> > Could you please prepare a short testcase, only containing the
> > necessary lines to reproduce the problem?
> 
> I have attached 2 files to reproduce the problem. It is linked to the
> putenv() call and the env variable then being reset.

Thanks for the testcase.

> 	strcpy(fooflags,"fooflags=1");
> 	putenv(fooflags);
> 	*fooflags = '\0'; /* remove from environment after all */

Oh well.  What happens is that your empty string results in a fatal
change in the Win32 environment propagated to the child application.
A Win32 environment consists of a long string of \0-terminated strings.
The end of the environment is indicated by a second \0, like this:

  "a=b\0c=d\0e=f\0\0"

By setting one of the strings to \0 as you do, the propagated environment
contains an early \0:

  "\0\0c=d\0e=f\0\0"

The result is that the propagated Win32 environment is actually empty.
Missing $PATH and $SYSTEMROOT are fatal and explain the socket problem
you encountered.

This is a bug in Cygwin insofar as Cygwin should not allow to propagate
empty strings to the child application.  I'm preparing a patch right now.
In the meantime, don't do that.  Replace

  *fooflags = '\0';

by

  unsetenv ("fooflags");

NB: The use of setenv/unsetenv is preferrable over using putenv anyway.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          mailto:cygwin@cygwin.com
Red Hat, Inc.

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/



More information about the Cygwin mailing list