This patch fixes the 'make check' failure that occurs on non-mule UNIX builds, most recently reported by Vin, and is based on Ben's observations. The major change is that instead of disabling the test, I turn EOL detection on for the life of the test. The reason is that the detection code is still present in these builds and a user could enable it, so it might as well be tested. Looking through my notes I've discovered that this patch is remarkably similar to a test patch on the subject that I produced a couple of years ago (see http://list-archive.xemacs.org/xemacs-beta/200404/msg00397.html). It's nice to see that my suspicions have been confirmed. lisp/ChangeLog addition: 2005-12-10 Malcolm Purvis <malcolmp(a)xemacs.org&gt; * code-init.el (set-eol-detection): Add `default-buffer-file-coding-system' to the list of coding systems that are set. src/ChangeLog addition: 2005-12-10 Malcolm Purvis <malcolmp(a)xemacs.org&gt; * tests.c (Ftest_data_format_conversion): Force EOL detection to on during the life of a test that needs it. xemacs-make-check source patch: Diff command: cvs -q diff -u Files affected: src/tests.c lisp/code-init.el Index: lisp/code-init.el =================================================================== RCS file: /pack/xemacscvs/XEmacs/xemacs/lisp/code-init.el,v retrieving revision 1.10 diff -u -r1.10 code-init.el --- lisp/code-init.el 2005/01/26 09:51:23 1.10 +++ lisp/code-init.el 2005/12/10 12:24:23 ＠＠ -60,7 +60,7 ＠＠ default, but this may change. NOTE: You *REALLY* should not turn off EOL detection on Windows! Your files will have lots of annoying ^M's in them if you do this." - (dolist (x '(buffer-file-coding-system-for-read + (dolist (x '(buffer-file-coding-system-for-read default-buffer-file-coding-system keyboard default-process-coding-system-read no-conversion-coding-system-mapping)) Index: src/tests.c =================================================================== RCS file: /pack/xemacscvs/XEmacs/xemacs/src/tests.c,v retrieving revision 1.11 diff -u -r1.11 tests.c --- src/tests.c 2005/04/08 23:11:33 1.11 +++ src/tests.c 2005/12/10 12:24:28 ＠＠ -35,7 +35,6 ＠＠ static Lisp_Object Vtest_function_list; - DEFUN ("test-data-format-conversion", Ftest_data_format_conversion, 0, 0, "", /* Test TO_EXTERNAL_FORMAT() and TO_INTERNAL_FORMAT() */ ＠＠ -357,12 +356,30 ＠＠ intern ("no-conversion-unix")); DFC_CHECK_DATA_COND_EOL_NUL (ptr, len, ext_unix, int_foo); - TO_INTERNAL_FORMAT (LISP_OPAQUE, opaque_dos, - LISP_BUFFER, Fcurrent_buffer(), - intern ("undecided")); - /* &&#### needs some 8-bit work here */ - DFC_CHECK_DATA (BUF_BYTE_ADDRESS (current_buffer, BUF_PT (current_buffer)), - sizeof (int_foo) - 1, int_foo); + { + /* + This test requires EOL detection to be enabled, which isn't by default + in all circumstances (in particular non-Mule on UNIX). Work around that + by enabling EOL detection for this test only. + */ + Lisp_Object eol_detection_enabled_p = intern("eol-detection-enabled-p"); + Lisp_Object set_eol_detection = intern("set-eol-detection"); + + Lisp_Object old_eol = Fsymbol_value(eol_detection_enabled_p); + + Fset(eol_detection_enabled_p, Qt); + call1(set_eol_detection, Qt); + + TO_INTERNAL_FORMAT (LISP_OPAQUE, opaque_dos, + LISP_BUFFER, Fcurrent_buffer(), + intern ("undecided")); + /* &&#### needs some 8-bit work here */ + DFC_CHECK_DATA (BUF_BYTE_ADDRESS (current_buffer, BUF_PT (current_buffer)), + sizeof (int_foo) - 1, int_foo); + + Fset(eol_detection_enabled_p, old_eol); + call1(set_eol_detection, old_eol); + } TO_INTERNAL_FORMAT (DATA, (ext_mac, sizeof (ext_mac) - 1), LISP_STRING, string, -- Malcolm Purvis <malcolmp(a)xemacs.org&gt;