On 08/22/2012 02:51 PM, daniel.ja...@t-online.de wrote: > Hi Stefano, > Hi Daniel. Note that I'm re-adding the mailing list in CC:, so that our discussion will be registered in the Automake bug tracker.
> here are the files you asked for. > Thanks! > .../automake-1.12.3/t/lex-clean-cxx.dir/parsefoo.cxx > I see this: ... 84: int isatty (int fd) { return 0; } 85: # define YYNEWLINE 10 86: yylex(){ 87: int nstr; extern int yyprevious; 88: #ifdef __cplusplus ... so my diagnosis was correct. And since Solaris 10 lex no longer has this issue (yay), I'd say we can safely ignore the failure -- it's not Automake fault if Solaris 9 lex cannot produce code working with C++ compilers. > .../automake-1.12.3/t/tests-environment-fd-redirect.dir/configure > And this shows you ksh is failing with this line (a line wholly under the control of Autoconf BTW): # Prefer a ksh shell builtin over an external printf program on Solaris, # but without wasting forks for bash or zsh. if test -z "$BASH_VERSION$ZSH_VERSION" \ && (test "X`print -r -- $as_echo`" = "X$as_echo") 2>/dev/null; then ... This is very strange. What happens if you issue this from your prompt? $ ksh -c 'echo "$ZSH_VERSION"; echo "$BASH_VERSION"' > > In the meantime I rerun the tests using GNU bash 4.2.0 instead of ksh as > a shell. The results were the same. > Yeah, that was to be expected, because the failures involving ksh (in the texi2dvi script and in the configure script of the test 'tests-environment-fd-redirect.sh') are not under the control of Automake. To solve your problems with texi2dvi, you might need to ask your sysadmin to change its shebang line to point to a bash shell instead. Thanks, Stefano