mirror/guile
mirror/guile
1
Fork 0
mirror of https://git.savannah.gnu.org/git/guile.git synced 2025-06-10 22:10:21 +02:00
Code Activity

(scm_putenv): Confine the putenv("NAME=") bit to mingw, use

Browse source 
putenv("NAME") as the fallback everywhere else.  In particular this is
needed for solaris 9.  Reported by Frank Storbeck.
This commit is contained in:
Kevin Ryde 2007-09-11 00:46:15 +00:00
parent bc4ee34e1d
commit 5b7ecf1b30
1 changed files with 38 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

48
libguile/posix.c
View file

@ -1,4 +1,4 @@
/* Copyright (C) 1995,1996,1997,1998,1999,2000,2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
/* Copyright (C) 1995,1996,1997,1998,1999,2000,2001, 2002, 2003, 2004, 2005, 2006, 2007 Free Software Foundation, Inc.
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
@ -1340,16 +1340,39 @@ SCM_DEFINE (scm_putenv, "putenv", 1, 0, 0,
if (strchr (c_str, '=') == NULL)
{
#ifdef HAVE_UNSETENV
/* No '=' in argument means we should remove the variable from
the environment. Not all putenvs understand this (for instance
FreeBSD 4.8 doesn't). To be safe, we do it explicitely using
unsetenv. */
/* We want no "=" in the argument to mean remove the variable from the
environment, but not all putenv()s understand this, for example
FreeBSD 4.8 doesn't. Getting it happening everywhere is a bit
painful. What unsetenv() exists, we use that, of course.
Traditionally putenv("NAME") removes a variable, for example that's
what we have to do on Solaris 9 (it doesn't have an unsetenv).
But on DOS and on that DOS overlay manager thing called W-whatever,
putenv("NAME=") must be used (it too doesn't have an unsetenv).
Supposedly on AIX a putenv("NAME") could cause a segfault, but also
supposedly AIX 5.3 and up has unsetenv() available so should be ok
with the latter there.
For the moment we hard code the DOS putenv("NAME=") style under
__MINGW32__ and do the traditional everywhere else. Such
system-name tests are bad, of course. It'd be possible to use a
configure test when doing a a native build. For example GNU R has
such a test (see R_PUTENV_AS_UNSETENV in
https://svn.r-project.org/R/trunk/m4/R.m4). But when cross
compiling there'd want to be a guess, one probably based on the
system name (ie. mingw or not), thus landing back in basically the
present hard-coded situation. Another possibility for a cross
build would be to try "NAME" then "NAME=" at runtime, if that's not
too much like overkill. */
#if HAVE_UNSETENV
/* when unsetenv() exists then we use it */
unsetenv (c_str);
free (c_str);
#else
/* On e.g. Win32 hosts putenv() called with 'name=' removes the
environment variable 'name'. */
#elif defined (__MINGW32__)
/* otherwise putenv("NAME=") on DOS */
int e;
size_t len = strlen (c_str);
char *ptr = scm_malloc (len + 2);
@ -1359,7 +1382,12 @@ SCM_DEFINE (scm_putenv, "putenv", 1, 0, 0,
e = errno; free (ptr); free (c_str); errno = e;
if (rv < 0)
SCM_SYSERROR;
#endif /* !HAVE_UNSETENV */
#else
/* otherwise traditional putenv("NAME") */
rv = putenv (c_str);
if (rv < 0)
SCM_SYSERROR;
#endif
}
else
{