Hi, Ralf Menzel <[EMAIL PROTECTED]> writes: > The failure of tests 33 and 34 seems to occur when the testsuite is > executed by /bin/sh instead of /bin/bash. > > BTW, is this the indented behaviour? > > --- snip --- > bash> ../configure > /dev/null ; grep '^SHELL' Makefile > SHELL = /bin/bash > bash> /bin/bash ../configure > /dev/null ; grep '^SHELL' Makefile > SHELL = /bin/sh > bash> /bin/sh ../configure > /dev/null ; grep '^SHELL' Makefile > SHELL = /bin/bash > --- snip --- > > Looks odd to me.
I tried to find out what is causing this 'inconsistency'. (Please keep in mind that I don't fully understand what I'm doing.) To me it looks like the configure script tests if its executing shell is 'compatible enough'. If this is not the case it looks for a better shell and records its name in CONFIG_SHELL. If the script is happy with the executing shell it doesn't touch the variable CONFIG_SHELL. It then assigns the variable SHELL the value of CONFIG_SHELL with /bin/sh as default. So, if the shell that executes the configure script is bash the varible SHELL will be set to /bin/sh. It tried the following patch. --- snip --- Index: lib/autoconf/general.m4 =================================================================== RCS file: /cvsroot/autoconf/autoconf/lib/autoconf/general.m4,v retrieving revision 1.862 diff -u -p -r1.862 general.m4 --- lib/autoconf/general.m4 24 May 2005 07:32:13 -0000 1.862 +++ lib/autoconf/general.m4 26 May 2005 17:05:48 -0000 @@ -398,7 +398,7 @@ cross_compiling=no subdirs= MFLAGS= MAKEFLAGS= -AC_SUBST([SHELL], [${CONFIG_SHELL-/bin/sh}])dnl +AC_SUBST([SHELL], [${CONFIG_SHELL-${SHELL-/bin/sh}}])dnl AC_SUBST([PATH_SEPARATOR])dnl # Maximum number of lines to put in a shell here document. --- snip --- This helped in my case. So you think it might be useful in general? Bye, Ralf