Re: [PATCH] getcwd-lgpl: fix m4 to match relaxed test for BSD

[Bruno Haible]

Hi Jim,

Getting back to this issue from last Monday:

> > If we fix Eric's typo, REPLACE_GETCWD will come out as 0 on OpenBSD and 
> > NetBSD.
> 
> Do you know how that happens, given the corrected code?

It's because of the details in m4/getcwd-abort-bug.m4.

> The part that is surprising is that they would have
> $gl_cv_func_getcwd_path_max = yes.
> That would mean that getcwd (NULL, 0) can succeed when called
> from a directory whose name is longer than PATH_MAX.

No, it doesn't succeed. But the logic at the end of m4/getcwd-abort-bug.m4
claims that this failure was a success.

> > Can packages that use gnulib's getcwd() module - including coreutils -
> > expect that in normal conditions (no EPERM anywhere)
> >      mkdir ("subdir1", 0700);
> >      chdir ("subdir1");
> >      mkdir ("subdir2", 0700);
> >      chdir ("subdir2");
> >      ...
> >      mkdir ("subdir1000", 0700);
> >      chdir ("subdir1000");
> >      dir = getcwd (NULL, 0);
> > produces a non-NULL dir (longer than PATH_MAX) ?
> 
> Yes, assuming that nothing else goes wrong.

Thanks, this is clear.

> > So what you are saying is that the unit test and the documentation are
> > correct, and only the getcwd*.m4 files need to be changed so as to activate
> > the replacement in lib/getcwd.c. This replacement will first call the
> > system's getcwd() function, so as to handle file names shorter than PATH_MAX
> > with inaccessible parents.
> >
> > Is that what you mean?
> 
> Yes.

OK, clear.

> I think Eric proposed some POSIX wording change that would
> allow gnulib's getcwd to be considered compliant.

OK, I try to avoid making a firm claim about what POSIX says in this area.
I'm applying this fix.


2011-11-27  Bruno Haible  <address@hidden>

        getcwd: Fix bug from 2011-08-17.
        * m4/getcwd.m4 (gl_FUNC_GETCWD): Set REPLACE_GETCWD to 1 only on
        platforms that need it.
        * m4/getcwd-abort-bug.m4 (gl_FUNC_GETCWD_ABORT_BUG): Consider a return
        code of 4 to be a failure, not a success. This ensures that
        REPLACE_GETCWD becomes 1 on OpenBSD 4.9 and NetBSD 5.1.

--- m4/getcwd-abort-bug.m4.orig Mon Nov 28 00:00:21 2011
+++ m4/getcwd-abort-bug.m4      Sun Nov 27 23:50:37 2011
@@ -1,4 +1,4 @@
-# serial 6
+# serial 7
 # Determine whether getcwd aborts when the length of the working directory
 # name is unusually large.  Any length between 4k and 16k trigger the bug
 # when using glibc-2.4.90-9 or older.
@@ -91,7 +91,8 @@
      results in a failed assertion.  */
   cwd = getcwd (NULL, 0);
   if (cwd == NULL)
-    fail = 4; /* getcwd failed.  This is ok, and expected.  */
+    fail = 4; /* getcwd failed: it refuses to return a string longer
+                 than PATH_MAX.  */
   free (cwd);
 
   /* Call rmdir first, in case the above chdir failed.  */
@@ -110,11 +111,18 @@
 }
           ]])],
     [gl_cv_func_getcwd_abort_bug=no],
-    dnl A "regular" nonzero return does not indicate this bug.
-    dnl An abort will provoke an exit code of something like 134 (128 + 6).
-    [test $? -gt 128 \
-      && gl_cv_func_getcwd_abort_bug=yes \
-      || gl_cv_func_getcwd_abort_bug=no],
+    [dnl An abort will provoke an exit code of something like 134 (128 + 6).
+     dnl An exit code of 4 can also occur (in OpenBSD 4.9, NetBSD 5.1 for
+     dnl example): getcwd (NULL, 0) fails rather than returning a string
+     dnl longer than PATH_MAX.  This may be POSIX compliant (in some
+     dnl interpretations of POSIX).  But gnulib's getcwd module wants to
+     dnl provide a non-NULL value in this case.
+     ret=$?
+     if test $ret -ge 128 || test $ret = 4; then
+       gl_cv_func_getcwd_abort_bug=yes
+     else
+       gl_cv_func_getcwd_abort_bug=no
+     fi],
     [gl_cv_func_getcwd_abort_bug=yes])
   ])
   AS_IF([test $gl_cv_func_getcwd_abort_bug = yes], [$1], [$2])
--- m4/getcwd.m4.orig   Mon Nov 28 00:00:21 2011
+++ m4/getcwd.m4        Sun Nov 27 23:37:22 2011
@@ -6,7 +6,7 @@
 # with or without modifications, as long as this notice is preserved.
 
 # Written by Paul Eggert.
-# serial 10
+# serial 11
 
 AC_DEFUN([gl_FUNC_GETCWD_NULL],
   [
@@ -139,12 +139,12 @@
       ;;
   esac
 
-  case 
$gl_cv_func_getcwd_null,$gl_cv_func_getcwd_posix_signature$gl_cv_func_getcwd_path_max,$gl_abort_bug
 in
-  *yes,yes,yes,no) ;;
-  *)
-    dnl Full replacement lib/getcwd.c, overrides LGPL replacement.
-    REPLACE_GETCWD=1;;
-  esac
+  if { case "$gl_cv_func_getcwd_null" in *yes) false;; *) true;; esac; } \
+     || test $gl_cv_func_getcwd_posix_signature != yes \
+     || test "$gl_cv_func_getcwd_path_max" != yes \
+     || test $gl_abort_bug = yes; then
+    REPLACE_GETCWD=1
+  fi
 ])
 
 # Prerequisites of lib/getcwd.c, when full replacement is in effect.
-- 
In memoriam George Moscone <http://en.wikipedia.org/wiki/George_Moscone>